HISTORY OF HOMOLOGICAL ALGEBRA Charles A. Weibel

advertisement
HISTORY OF HOMOLOGICAL ALGEBRA
Charles A. Weibel
Homological algebra had its origins in the 19th century, via the work of Riemann
(1857) and Betti (1871) on “homology numbers,” and the rigorous development of
the notion of homology numbers by Poincaré in 1895. A 1925 observation of Emmy
Noether [N25] shifted the attention to the “homology groups” of a space, and
algebraic techniques were developed for computational purposes in the 1930’s. Yet
homology remained a part of the realm of topology until about 1945.
During the period 1940-1955, these topologically-motivated techniques for computing homology were applied to define and explore the homology and cohomology
of several algebraic systems: Tor and Ext for abelian groups, homology and cohomology of groups and Lie algebras, and the cohomology of associative algebras. In
addition, Leray introduced sheaves, sheaf cohomology and spectral sequences.
At this point Cartan and Eilenberg’s book [CE] crystallized and redirected the
field completely. Their systematic use of derived functors, defined via projective
and injective resolutions of modules, united all the previously disparate homology
theories. It was a true revolution in mathematics, and as such it was also a new
beginning. The search for a general setting for derived functors led to the notion
of abelian categories, and the search for nontrivial examples of projective modules
led to the rise of algebraic K-theory. Homological algebra was here to stay.
Several new fields of study grew out of the Cartan-Eilenberg Revolution. The
importance of regular local rings in algebra grew out of results obtained by homological methods in the late 1950’s. The study of injective resolutions led to
Grothendieck’s theory of sheaf cohomology, the discovery of Gorenstein rings and
Local Duality in both ring theory and algebraic geometry. In turn, cohomological
methods played a key role in Grothendieck’s rewriting of the foundations of algebraic geometry, including the development of derived categories. Number theory
was infused with new results from Galois cohomology, which in turn led to the
development of étale cohomology and the eventual solution of the Weil Conjectures
by Deligne.
Simplicial methods were introduced in the 1950’s by Kan, Dold and Puppe.
They led to the rise of homotopical algebra and nonabelian derived functors in the
1960’s. Among its many applications, perhaps André-Quillen homology for commutative rings and higher algebraic K-theory are the most noteworthy. Simplicial
methods also played a more recent role in the development of Hochschild homology,
topological Hochschild homology and cyclic homology.
This completes a quick overview of the history we shall discuss in this article.
Now let us turn to the beginnings of the subject.
Typeset by AMS-TEX
2
CHARLES A. WEIBEL
Betti numbers, Torsion Coefficients and the rise of Homology
Homological algebra in the 19th century largely consisted of a gradual effort to
define the “Betti numbers” of a (piecewise linear) manifold. Beginning with Riemann’s notion of genus, we see the gradual development of numerical invariants
by Riemann, Betti and Poincaré: the Betti numbers and Torsion coefficients of a
topological space. Indeed, the subject did not really move beyond these numerical invariants until about 1930. And it was not concerned with anything except
invariants of topological spaces until about 1945.
Riemann and Betti.
The first step was taken by Riemann (1826–1866) in his great 1857 work “Theorie
der Abel’schen Funktionen” [Riem, VI]. Let C be a system of
R one or more closed
curves Cj on a surface S, and consider the contour integral C X dx + Y dy of an
exact differential form. Riemann remarked that this integral vanished if C formed
the complete boundary of a region in S (Stokes’ Theorem), and this led him to
a discussion of “connectedness numbers.” Riemann defined S to be (n + 1)-fold
connected if there exists a family C of n closed curves Cj on S such that no subset
of C forms the complete boundary of a part of S, and C is maximal with this
property. For example, S is “simply connected” (in the modern sense) if it is 1-fold
connected. As we shall see, the connectness number of S is the homology invariant
1 + dim H1 (S; Z/2).
Riemann showed that the connectedness number of S was independent of the
choice of maximal family C. The key to his assertion is the following result, which
is often called “Riemann’s Lemma” [Riem, p. 85]: Suppose that A, B and C are
three families of curves on S such that A and B form the complete boundary of
one region of S, and A and C form the complete boundary of a second region of S.
Then B and C together must also form the boundary of a third region, obtained as
the symmetric difference of the other two regions (obtained by adding the regions
together, and then subtracting any part where they overlap).
If we write C ∼ 0 to indicate that C is a boundary of a region then Riemann’s
Lemma says that if A + B ∼ 0 and A + C ∼ 0 then B + C ∼ 0. This, in modern
terms, is the definition of addition in mod 2 homology! Indeed, the Cj in a maximal
system form a basis of the singular homology group H1 (S; Z/2).
Riemann was somewhat vague about what he meant by “closed curve” and “surface,” but we must remember that this paper was written before Möbius discovered
the “Möbius surface” (1858) or Peano studied pathological curves (1890). There is
another ambiguity in this Lemma, pointed out by Tonelli in 1875: every curve Cj
must be used exactly once.
Riemann also considered the effect of making cuts (Querschnitte) in S. By
making each cut qj transverse to a curve Cj (see [Riem, p. 89]), he showed that
the number of cuts needed to make S simply connected equals the connectivity
number. For a compact Riemann surface, he shows [Riem, p. 97] that one needs
an even number 2p of cuts. In modern language, p is the genus of S, and the
interaction between the curves Cj and cuts qj forms the germ of Poincaré Duality
for H1 (S; Z/2).
Riemann had poor health, and frequently visited Italy for convalescence between
1858 and his death in 1866. He frequently visited Enrico Betti (1823–1892) in
HISTORY OF HOMOLOGICAL ALGEBRA
3
Pisa, and the two of them apparently discussed the idea of extending Riemann’s
construction to higher dimensional manifolds. Two documents with very similar
definitions survive.
One is an undated “Fragment on Analysis Situs” [Riem, XXVIII], discovered
among Riemann’s effects, in which Riemann defines the n-dimensional connectedness of a manifold M : replace “closed curve” with n-dimensional subcomplex
(Streck) without boundary, and “bounding a region” with “bounding an (n + 1)dimensional subcomplex). Riemann also defined higher dimensional cuts (submanifolds whose boundary lies on the boundary of M ) and observed that a cut of dimension dim(M ) − n either drops the n-dimensional connectivity by one, or raises
the (n − 1)-dimensional connectivity by one. In fairness, we should point out that
Riemann’s notion of connectedness is not independent of the choice of basis, because his notion that A and B are similar (veränderlich) is not the same as A and
B being homologous; a counterexample was discovered by Heegard in 1898.
The other document is Betti’s 1871 paper [Betti]. The ideas underlying this
paper are the same as those in Riemann’s fragment, and Betti states that his proof
of the independence of the homology numbers from the choice of basis is based
upon the proof in Riemann’s 1857 paper. However, Heegard observed in 1898 that
Betti’s proof of independence is not correct in several respects, starting from the
fact that a meridian on a torus is not closed in Betti’s sense.
Betti also made the following assertion ([Betti, p.148]), which presages the
Poincaré Duality Theorem: “In order to render a finite n-dimensional space simply
connected, by removing simply connected sections, it is necessary and sufficient
to make pn−1 linear cuts, . . . , p1 cuts of dimension n − 1,” where pi + 1 is the
ith connectivity number. Heegard found mistakes in Betti’s proof here too, and
Poincaré observed in 1899 [Poin, p. 289] that the problem was in (Riemann and)
Betti’s definition of similarity: it is not enough to just consider the set underlying
A, one must also account for multiplicities.
Poincaré and Analysis Situs.
Inspired by Betti’s paper, Poincaré (1854–1912) developed a more correct homology theory in his landmark 1895 paper “Analysis Situs” [Poin]. After defining the
notion, he fixes a piecewise linear manifold (variété) V . Then he considers formal
integer combinations of oriented n-dimensional submanifolds Vi of V , and introduces a relation called a homology, which can be added like ordinary equations:
P
ki Vi ∼ 0 if there is an (n + 1)-dimensional submanifold W whose boundary
consists of k1 submanifolds like V1 , k2 submanifolds like V2 , etc.
Poincaré calls a family of n-dimensional submanifolds Vi linearly independent
if there is no homology (with integer coefficients) connecting them. In honor of
Enrico Betti, Poincaré defined the nth Betti number of V to be bn + 1, where bn is
the size of a maximal independent family. Today we call bn the nth Betti number,
because it is the dimension of the rational vector space Hn (V ; Q). For geometric
reasons, he did not bother to define the nth Betti number for n = 0 or n = dim(V ).
With this definition, Poincaré stated his famous Duality Theorem [Poin, p. 228]:
for a closed oriented (m-dimensional) manifold, the Betti numbers equally distant
from the extremes are equal, viz., bi = bm−i . Unfortunately, there was a gap
in Poincaré’s argument, found by Heegard in 1898. Poincaré published a new
4
CHARLES A. WEIBEL
P
proof in 1899, using a triangulation of V and restricting his formal sums
ki Vi to
linear combinations of the simplices in the triangulation. Of course this restriction
yields “reduced” Betti numbers which could potentially be different than the Betti
numbers he had defined in 1895. Using simplicial subdivisions, he sketched a proof
that these two kinds of Betti numbers agreed. (His sketch had a geometric gap,
which was filled in by J. W. Alexander in 1915.) This 1899 paper was the origin of
the simplicial homology of a triangulated manifold.
Poincaré’s 1899 paper also contains the first appearance of what would eventually
(after 1929) be called a chain complex. Let V be an oriented polyhedron. On p. 295
of [Poin], he defined boundary matrices εq as follows. The (i, j) entry describes
whether or not the jth (q − 1)-dimensional simplex in V lies on the boundary of the
ith q-dimensional simplex: εqij = ±1 if it is (+1 if the orientation is the same, −1 if
not) and εqij = 0 if they don’t meet. Poincaré called the collection of these matrices
the scheme of the polyhedron, and demonstrated on p. 296 that εq−1 ◦ εq = 0. This
is of course the familiar condition that the matrices εq form the maps in a chain
complex, and today Poincaré’s scheme is called the simplicial chain complex of the
oriented polyhedron V .
Another major result in Analysis Situs is the generalization of the notion of
Euler characteristic to higher dimensional polyhedra V . If αn is the P
number of
n-dimensional cells, Poincaré showed that the alternating sum χ(V ) = (−1)n αn
is independent of the choice of triangulation of V (modulo the gap filled by Alexander). On p. 288 he showed that χ(V ) is the alternating sum of the Betti numbers
bn (in the modern sense); because of this result χ(V ) is today called the EulerPoincaré characteristic of V . Finally, when V is closed and dim(V ) is odd, he used
Duality to deduce that χ(V ) = 0.
In 1900, Poincaré returned once again to the subject of homology, in the Second complément à l’Analysis Situs. This paper is important from our perspective
because it introduced linear algebra and the notion of torsion coefficients. To do
this, Poincaré considered the sequence of integer matrices (or tableaux) T p which
describe the boundaries of the p-simplices in a polyhedron; this sequential display
of integer matrices was the second occurrence of the notion of chain complex.
In Poincaré’s framework, one performs elementary row and column operations
upon the all the matrices until the matrix Tp had been reduced to the block form

I

Tp = 0
0
0
Kp
0

0
0,
0

Kp = 

k1
k2
..
.
,
1 < k1 , k1 |k2 , k2 |k3 , · · ·
Here I denotes an identity matrix. The pth Betti number bp is the difference
between the number of zero columns in Tp and the number of nonzero rows in Tp+1
[Poin, p. 349]. The pth torsion coefficients were defined as the integers k1 , k2 ,
etc. in the matrix Kp+1 [Poin, p. 363].
In modern language, Hn (V ; Z) is a finitely generated abelian group, so it has the
form Zbn ⊕ Z/k1 ⊕ Z/k2 ⊕ · · · with k1 |k2 , k2 |k3 , etc. Here bn is the Betti number,
and the pth torsion coefficients are the orders of the finite cyclic groups Z/ki . Of
course, since homology was not thought of as a group until 1925 (see [N25]), this
formulation would have looked quite strange to Poincaré!
HISTORY OF HOMOLOGICAL ALGEBRA
5
Homology of topological spaces (1900–1935).
The next 25 years were a period of consolidation and clarification of Poincaré’s
ideas. For example, the Duality Theorem for the mod 2 Betti numbers, even for
nonoriented manifolds, appeared in the 1913 paper [VA] by O. Veblen (1880–1960)
and J. W. Alexander (1888–1971). The topological invariance of the Betti numbers
and torsion coefficients of a manifold was established by Alexander in 1915. In
1923, Hermann Künneth (1892–1975) calculated the Betti numbers and torsion
coefficients for a product of manifolds in [K23]; his results have since become known
as the Künneth Formulas.
Until the mid 1920’s, topologists studied homology via incidence matrices, which
they could manipulate to determine the Betti numbers and torsion coefficients.
This changed in 1925, when Emmy Noether (1882–1935) pointed out in her 14-line
report [N25], and in her lectures in Göttingen, that homology was an abelian group,
rather than just Betti numbers and torsion coefficients, and perceptions changed
forever. The young H. Hopf (1894–1971), who had just arrived to spend a year in
Göttingen and meet P. Alexandroff, realized how useful this viewpoint was, and
the word spread rapidly. Inspired by the new viewpoint, the 1929 paper [M29] by
L. Mayer (1887–1948) introduced the purely algebraic notions of chain complex,
its subgroup of cycles and the homology groups of a complex. Slowly the subject
became more algebraic.
During the decade 1925-1935 there was a general movement to extend the principal theorems of algebraic topology to more general spaces than those considered
by Poincaré. This led to several versions of homology. Some people who invented
homology theories in this decade were: Alexander [A26], Alexandroff (1896–1982),
Čech (1893–1960) [C32], Lefschetz (1884–1972) [L33], Kolmogoroff (1903–1987),
Kurosh (1908–1971) and Vietoris (1891–!). In 1940, Steenrod (1910–1989) developed a homology theory for compact metric spaces [S40], and his theory also belongs
to this movement.
In each case, the homology theory could be described as follows: given topological
data, the inventors gave an ad hoc recipe for constructing a chain complex, and
defined their homology groups to be the homology of that chain complex. In each
case, they showed that the result is independent of choices, and provided the usual
Betti numbers for compact manifolds. One theme in many recipes was homology
with coefficients in a compact topological group; this kind of homology remained
in vogue until the early 1950’s, by which time it had become superfluous. We
shall pass over most of this decade, as it played little part in the development of
homological algebra per se.
One theory we should mention is the “de Rham homology” of a smooth manifold,
which was introduced by G. de Rham (1903–1990) in his 1931 thesis [dR]. Elie
Cartan (1869–1951) had just introduced the cochain complex of exterior differential
forms on a smooth manifold M in a series of papers [C28, C29] and had conjectured
that the Betti number bi of M is the
Pmaximum number of closed i-forms ωj such
that no nonzero linear combination
λj ωj is exact. When de Rham saw Cartan’s
note [C28] in 1929, he quickly realized that he could solve Cartan’s conjecture using
a triangulation on M and the bilinear map
Z
(C, ω) 7→
ω.
C
6
CHARLES A. WEIBEL
Here C is an i-cycleRfor the triangulation and ω is a closed i-form. Indeed, Stokes’
formula shows that C ω = 0 if either ω is an exact form
R or if C is a boundary. De
Rham showed the converse was true:
R if we fix C then C ω = 0 if and only if C is
a boundary, while if we fix ω then C ω = 0 if and only if ω is exact. De Rham’s
i
theorem proves Cartan’s conjecture, since if we write HdR
(M ) for the quotient of
all closed forms by the exact forms, then it gives a nondegenerate pairing between
i
the vector spaces Hi (M ; R) and HdR
(M ).
i
Of course, HdR (M ) is just the ith cohomology of Cartan’s complex, and we now
refer to it as the “de Rham cohomology” of M . But cohomology had not been
invented in 1931, and no one seems to have realized this fact until Cartan and
Chevalley in the 1940’s, so de Rham was forced to state his results in terms of
homology. Much later, the de Rham cohomology of Lie groups would then play a
critical role in the development of the cohomology of Lie algebras (see [ChE] and
the discussion below).
The rise of algebraic methods (1935–1945).
The year 1935 was a watershed year for topology in many ways. We shall focus
upon four developments of importance to homology theory.
The Hurewicz maps h : πn (X) → Hn (X; Z) were constructed and studied by
Witold Hurewicz (1904–1956) in 1935. Hurewicz also studied aspherical spaces,
meaning spaces such that πn (X) = 0 for n 6= 1. He noticed in [Hu36] that if X and
X 0 are two finite dimensional aspherical spaces with π1 (X) = π1 (X 0 ) then X and
X 0 are homotopy equivalent. From this he concluded that the homology Hn (X; Z)
of such an X depended only upon its fundamental group π1 (X). This observation
forms the implicit definition of the cohomology of a group, a definition only made
explicit a decade later (see below).
The homology of the classical Lie groups was calculated in 1935 by Pontrjagin
[P35] (Betti numbers only, using combinatorial proofs) and more fully by R. Brauer
[B35] (ring structure, using de Rham homology). These calculations led directly
to the modern study of Hopf algebras, as follows. In 1941, H. Hopf introduced
H-spaces in [Hf41], and showed that the Brauer-Pontrjagin calculations were a
consequence of the fact that the cohomology ring H ∗ (M ; Q) of any H-space M
is an exterior algebra on odd generators; today we would say that Hopf’s result
amounted to an early classification of finite-dimensional graded “Hopf algebras”
over Q.
The third major advance was the determination of Universal Coefficient groups
for homology, that is, a coefficient group Au which would determine the homology
groups H∗ (X; A) for arbitrary coefficients A. For finite complexes, where matrix
methods apply, J. W. Alexander had already shown in 1926 [A26] that H∗ (X; Z/n)
was determined by H∗ (X; Z), the case n = 2 having been done as early as 1912
[VA]. In the 1935 paper [C35], E. Čech discovered that Z is a Universal Coefficient
group for homology: assume that there is a chain complex C∗ of free abelian groups,
whose homology gives the integral homology of a space X (the space is introduced
only for psychological reasons). Then for every abelian group A and every complex
X, Hn (X; A) is the direct sum of two subgroups, determined explicitly by Hn (X; Z)
and Hn−1 (X; Z), respectively.
In fact, Čech’s Universal Coefficient Theorem gave explicit presentations for
HISTORY OF HOMOLOGICAL ALGEBRA
7
these subgroups, which today we would recognise as presentations for Hn (X; Z) ⊗
A and Tor1 (Hn−1 (X; Z), A). Thus Čech was the first to introduce the general
tensor product and torsion product Tor of abelian group into homological algebra.
However, such a modern formulation of Čech’s result (and the name Tor, due to
Eilenberg around 1950) did not appear in print before 1951 (Exposé 10 of [C50]; see
also p. 161 of [ES]). We note a contemporary variant in passing: Steenrod proved
a Universal Coefficient Theorem for cohomology with coefficients in a compact
topological group; see [S36]; in this context the Universal Coefficient group is the
character group R/Z of Z.
The fourth great advance in 1935 was the discovery of cohomology theory and
cup products, simultaneously and independently by Alexander and Kolmogoroff.
The drama of their back-to-back presentations at the Moscow International Conference on Topology in September 1935 is nicely described in Massey’s article [MHC]
in this book. The Alexander-Kolmogoroff formulas defining the cup product were
completely ad hoc, and also not exactly correct; the rectification was quickly discovered by Čech and Hassler Whitney (1907–1989), and corrected by Alexander.
All three authors published articles about the cup product in the Annals of Mathematics during 1936–1938. Whitney’s article [W37] had the most enduring impact,
for it introduced the modern “co” terminology: coboundary (δ) and cocycle, as well
as the notation a ` b and a a b, prophetically suggesting that “we might call `
‘cup’ and a ‘cap’.” Whitney’s article also implicitly introduced the notion of what
we now call a differential graded algebra, via the “Leibniz axiom” that if a and b
are homogeneous of degrees p and q then:
δ(a ` b) = (δa) ` b + (−1)p a ` δb.
During the next decade, while the world was at war, the algebraic machinery
slowly fell into place.
In the 1938 paper [W38], Hassler Whitney discovered the tensor product construction A ⊗ B for abelian groups (and modules). Up to that time, this operation
had only been known (indirectly) in special cases: the tensor product of vector
spaces, or the tensor product of A with a finitely generated abelian group B. Whitney took the name from the following classical example in differential geometry: if
T is the tangent vector space of a manifold at a point, then T ⊗T is the vector space
of (covariant) “tensors of order 2.” The full modern definition of the tensor product
(using left and right modules) appeared in Bourbaki’s influential 1943 treatment
[B43], as well as in the 1944 book [ANT] by Artin, Nesbitt and Thrall.
The concept of an exact sequence first appeared in Hurewicz’ short abstract
[Hu41] of a talk in 1941. This abstract discusses the long exact sequence in cohomology associated to a closed subset Y ⊂ X, in which the operation δ : H q (X − Y ) →
H q+1 (X, Y ) plays a key role.
In the 1942 paper [EM42], Eilenberg (1915–) and Mac Lane (1909–) gave a treatment of the Universal Coefficient Problem for cohomology, naming Hom and Ext
for the first time. Using these, they showed that Čech homology with coefficients
in any abelian group A are determined by Čech cohomology with coefficients in Z.
This application further established the importance of algebra in topology. We will
say more about this discovery in the next section.
8
CHARLES A. WEIBEL
In 1944, S. Eilenberg defined singular homology and cohomology in [E44]. First,
he introduced the singular chain complex S(X) of a topological space, and then
he defined H∗ (X; A) and H ∗ (X; A) to be the homology and cohomology of the
chain complexes S(X) ⊗ A and Hom(S(X), A), respectively. The algebra of chain
complexes was now firmly entrenched in topology. Eilenberg’s definition of S(X)
was only a minor modification of Lefschetz’ construction in [L33], replacing the
notion of oriented simplices by the use of simplices with ordered vertices; this trick
avoided the issue of equivalence relations on oriented simplices which introduced
“degenerate” chains of order 2. (See [MHC].)
We close our description of this era with the 1945 paper by Eilenberg and Steenrod’s [ES45]. This paper outlined an axiomatic treatment of homology theory, rederiving the whole of homology theory for finite complexes from these axioms. They
also pointed out that singular homology and Čech homology satisfy the axioms, so
they must agree on all finite complexes. The now-familiar axioms introduced in
this paper were: functoriality of Hq and ∂; homotopy invariance; long exact homology sequence for Y ⊂ X; excision; and the dimension axiom: if P is a point
then Hq (P ) = 0 for q 6= 0. We refer the reader to the article [May] for subsequent
developments on generalized homology theories, which are characterized by the
Eilenberg-Steenrod axioms with the dimension axiom replaced by Milnor’s wedge
axiom [M62].
Homology and cohomology of algebraic systems
During the period 1940–1950, topologists gradually began to realize that the
homology theory of topological spaces gave invariants of algebraic systems. This
process began with the discovery that group extensions came up naturally in cohomology. Then came the discovery that the cohomology of an aspherical space Y
and of a Lie group G only depended upon algebraic data: the fundamental group
π = π1 (Y ) and the Lie algebra g associated to G, respectively. This led to thinking
of the homology and cohomology groups of Y and G as intrinsic to π and g, and
therefore algebraically defineable in terms of the group π and the Lie algebra g.
Ext of abelian groups.
If A and B are abelian groups, an extension of B by A is an abelian group E,
containing B as a subgroup, together with an identification of A with E/B. The
set Ext(A, B) of (equivalence classes of) extensions appeared as a purely algebraic
object, as a special case of the more general problem of group extensions (see below),
decades before it played a crucial role in the development of homological algebra.
Here is the approach used by Reinhold Baer in 1934 [B34]. Suppose that we fix
a presentation of an abelian group A by generators and relations: write A = F/R,
where F is a free abelian group, say with generators {ei }, and R is the subgroup
of relations. If E is any extension of B by A, then by lifting the generators of A to
elements a(ei ) of E we get an element a(r) of B for every relation r in R. Brauer
thought of this as a function from the defining relations of A into B, so he called the
induced homomorphism a : R → B a relations function. Conversely, he observed
that every relations function a gives rise to a factor set, and hence to an extension
E(a), showing that two relations functions a and a0 gave the same extensions if
and only if there are elements bi (corresponding to a function b : F → B) so that
HISTORY OF HOMOLOGICAL ALGEBRA
9
a0 (r) = b(r) + a(r) [B34, p. 394]. Finally, Baer observed (p. 395) that the formal
sum a + a0 of two relations functions defined an addition law on Ext(A, B), making
it into an abelian group. In his honor, we now call the extension E(a + a0 ) the
“Baer sum” of the extensions.
Baer’s presentation A = F/R amounted to a free resolution of A, and his formulas were equivalent to the modern calculation of Ext(A, B) as the cokernel of
Hom(F, B) → Hom(R, B). But working with free resolutions was still a decade
away ([Hf44, F46]), and using them to calculate Ext(A, B) was even further in the
future ([ES]).
We now turn to 1941. That year, Saunders Mac Lane gave a series of lectures
on group extensions at the University of Michigan. According to [M88], most of
the lectures concerned applications to Galois groups and class field theory, but
Mac Lane ended with a calculation of the abelian extensions of Z by A = Z[ p1 ].
Samuel Eilenberg, who had recently emmigrated from Poland and was an Instructor
at Michigan, could not attend the last lecture and asked for a private lecture.
Eilenberg immediately noticed that the group Z[ p1 ] was dual to the topological padic solenoid group Σ, which Eilenberg had been studying, and that Mac Lane’s
algebraic answer Ext(Z[ p1 ], Z) = Ẑp /Z coincided with the Steenrod homology groups
H1 (S 3 − Σ; Z) calculated (by Steenrod) in [S40]. After an all-night session, followed
by several months of puzzling over this observation, they figured out how Ext plays
a role in cohomology; the result was the paper [EM42].
Time has recognized their result as the Universal Coefficient Theorem for singular
cohomology. However, singular cohomology had not yet been invented in 1942. In
addition, the notation then in vogue, and used in [EM42], was the opposite of
today’s conventions (which date to [ES45] and [CE]) in several respects. They
wrote H q (A) for the homology groups they worked with, and wrote Hq (A) for the
cohomology groups under consideration. In addition, since they were reworking
many of Baer’s observations about extensions, they wrote Ext{B, A} for what we
call Ext(A, B).
Here is a translation of their Universal Coefficient Theorem into modern language. Given an infinite but star-finite CW complex K, they formed the cochain
complex C ∗ (K) of finite cocycles with integer coefficients; each C q (K) is a free
abelian group. Define the cohomology H∗ of K using C ∗ (K), and define the homology H∗ (K; A) of K with coefficients in A using the chain complex Hom(C ∗ (K), A).
Then Hq (K; A) is the product of Hom(Hq , A) and the group Ext(Hq+1 , A) of
abelian group extensions. Of course, the proof in [EM42] only uses the algebraic properties of C ∗ (K). Since [ES] and [CE] it has been traditional to state
this result the other way: given a chain complex C∗ of free abelian groups, one sets
Hq = Hq (C∗ ) and describes the cohomology of the cochain complex Hom(C∗ (K), A)
as the product of Hom(Hq , A) and Ext(Hq−1 , A).
In order to find a universal coefficient formula for the cohomology H q (K; A)
of C ∗ (K) ⊗ A, they discovered the “adjunction” isomorphism Hom(A ⊗ B, C) ∼
=
Hom(B, Hom(A, C)); see [EM42, (18.3)]. This is an isomorphism which varies
naturally with the abelian groups A, B and C. With this in hand, they reformulated
Čech’s Universal Coefficient Theorem: Hq (K; A) is the direct sum of Hq ⊗ A and a
group T that we would write as T = Homcont (A∗ , Hq−1 ), where A∗ is the Pontrjagin
10
CHARLES A. WEIBEL
dual of A. In fact, T is Tor(A, Hq−1 ); see [CE, p. 138].
The notion that Hom(A, B) varies naturally, contravariantly in A and covariantly
in B, was central to the discussion in [EM42]. In order to have a precise language
for speaking of this property for Hom, and for homology and cohomology, Eilenberg
and Mac Lane concocted the notions of functor and natural isomorphism in 1942.
They expanded the language to include category and natural transformation in
1945; see [EM45]. Although these concepts were used in several papers, the new
language of Category Theory did not gain wide acceptance until the appearance of
the books [ES] and [CE] in the 1950’s.
Cohomology of Groups.
The low dimensional cohomology of a group π was classically studied in other
guises, long before the notion of group cohomology was formulated in 1943–45. For
example, H 0 (π; A) = AG , H1 (π; Z) = π/[π, π] and (for π finite) the character group
H 2 (π, Z) = H 1 (π; C∗ ) = Hom(π, C∗ ) were classical objects.
The group H 1 (π, A) of crossed homomorphisms of π into a representation A is
just as classical: Hilbert’s “Theorem 90” (1897) is actually the calculation that
H 1 (π, L× ) = 0 when π is the Galois group of a cyclic field extension L/K, and the
name comes from its role in the study of crossed product algebras [BN].
The study of H 2 (π; A), which classifies extensions over π with normal subgroup
A via factor sets, is equally venerable. The idea of factor sets appeared as early as
Hölder’s 1893 paper [Hö, §18], again in Schur’s 1904 study [S04] of projective representations π → P GLn (C) (these determine an extension E over π with subgroup
C∗ , equipped with an n-dimensional representation) and again in 1906 in Dickson’s
construction of crossed product algebras. The first systematic treatment of factor
sets was O. Schrier’s 1926 paper [S26]; Schrier did not asssume that A was abelian.
In 1928, R. Brauer used factor sets in [B28] to represent central simple algebras
as crossed product algebras in relation to the Brauer group; this was clarified in
[BN]. In 1934, R. Baer gave the first invariant treatment of extensions (i.e., without
using factor sets) in [B34]. He noticed that when A was abelian, Schrier’s factor
sets could be added termwise, so that the extensions formed an abelian group.
Extensions with A abelian were also studied by Marshall Hall in [H38].
The next step came in 1941, when Heinz Hopf submitted a surprising 2-page
announcement [Hf41] to a topology conference at the University of Michigan. In it
he showed that the fundamental group π = π1 (X) determined the cokernel of the
Hurewicz map h : π2 (X) → H2 (X; Z). If we present π as the quotient π = F/R of
a free group F by the subgroup R of relations, Hopf gave the explicit formula:
H2 (X; Z) ∼ R ∩ [F, F ]
=
.
[F, R]
h π2 (X)
In particular, if π2 (X) = 0 this shows exactly how H2 (X; Z) depends only upon
π1 (X); this formula is now called Hopf ’s formula for H2 (π; Z).
Communication with Switzerland was difficult during World War II, and Hopf’s
paper arrived too late to be presented at the conference, but his result made a
big impression upon Eilenberg. According to Mac Lane [M88], Eilenberg suggested
that they try to get rid of that non-invariant presentation of π(X). Since they had
HISTORY OF HOMOLOGICAL ALGEBRA
11
just learned in [EM42] that homology determined cohomology, was it more efficient
to describe the effect of π1 (X) on H 2 (X; Z)? Mac Lane states that this line of
investigation provided the justification for the abstract study of the cohomology of
groups, and “was the starting point of homological algebra” ([ML, p. 137]).
The actual definition of the homology and cohomology of a group π first appeared
in the announcement [EM43] by Eilenberg and Mac Lane (the full paper appeared
in 1945). At this time (March 1943 until 1945) Eilenberg and Mac Lane were housed
together at Columbia, working on war-related applied mathematics [M89]. Independently in Amsterdam, Hans Freudenthal (1905-1990) discovered homology and
cohomology of groups using free resolutions; his paper [F46] was probably smuggled out of the Netherlands in late 1944. Also working independently of EilenbergMac Lane and Freudenthal, but in Switzerland, homology was defined in Hopf’s
paper [Hf44], and (based on Hopf’s paper) the cohomology ring was defined in
Beno Eckmann’s 1945 paper [Eck]. We will discuss these approaches, beginning
with [EM43].
Given π, Eilenberg and Mac Lane choose an aspherical space Y with π = π1 (Y ).
Using Hurewicz’ observation that the homology and cohomology groups of Y (with
coefficients in A) were independent of the choice of Y , Eilenberg and Mac Lane took
them as the definition of Hn (π; A) and H n (π; A). To perform computations, Eilenberg and Mac Lane chose a specific abstract simplicial complex K(π) for the aspherical space Y . Its n-cells correspond to ordered arrays [x1 , · · · , xn ] of elements
in the group. Thus one way to calculate the cohomology groups of π was to use the
cellular cochain complex of K(π), whose n-chains are functions f : π q → A from q
copies of π to A. Eckmann’s paper [Eck] also defines H q (X; A) as the cohomology
of this ad hoc cochain complex, and defines the cohomology cup product in terms
of this complex. Both papers showed that H 2 (G; A) classifies group extensions.
At the same time, Hopf gave a completely different definition in [Hf44]. First
Hopf considers a module M over any ring R, and constructs a resolution F∗ of M
by free R-modules. If I is an ideal of R, he considers the homology of the kernel of
F∗ → F∗ /I and shows that it is independent of the choice of resolution. In effect,
this is the modern definition of the groups TorR
∗ (M, R/I)! Hopf then specializes to
the group ring R = Z[π], the augmentation ideal I and M = Z, and defines the
homology of π to be the result. That is, Hopf’s definition is literally (in modern
notation)
Hn (π; Z) = TornZ[π] (Z, Z).
Finally, Hopf showed that if Y is an aspherical cell space with π = π1 (Y ) then
Hn (Y ; Z) = Hn (π; Z). His proof has since become standard: the cellular chain
complex F∗ for the universal cover of Y is a free Z[π]-resolution of Z, and that
F∗ /I is the cellular chain complex of Y . Thus the homology of F∗ /I simultaneously
computes the Betti homology of Y and the group homology of π, as claimed.
Freudenthal’s method [F46] was similar to Hopf’s, but less general. He considered
a free Z[π]-module resolution F∗ of Z, and showed that the homology of F∗ ⊗π
A is independent of F∗ for every abelian group A. Like the others, Freudenthal
constructed one such resolution from an aspherical polytope Y with π = π1 (Y ).
At first, calculations of group homology were restricted to those groups π which
were fundamental groups of familiar topological spaces, using the bar complex. In
12
CHARLES A. WEIBEL
his 1946 Harvard thesis [Lyn], R. Lyndon found a way to calculate the cohomology of a group π, given a normal subgroup N such that H ∗ (N ) and H ∗ (π/N ; A)
were known. His procedure started with H p (π/N ; H q (N )) and proceeded through
successive subquotients, ending with graded groups associated to a filtration on
H ∗ (π). Serre quickly realized [S50] that Lyndon’s procedure amounted to a spectral sequence, and completed the description with Hochschild in [HS53]. Since then,
it has been known as the Lyndon-Hochschild-Serre spectral sequence.
One application of the new definitions was Galois cohomology, so named in
Hochschild’s study [Hh50] of local class field theory. If L is a finite Galois extension of a field K with Galois group G, this referred to the cohomology of G
with coefficients in L× , or in a related G-module such as the idèle class group of
L. For example, the Normal Basis Theorem implies that the additive group L is a
free G-module over L, so H q (G; L) = 0 for q 6= 0 [E49]. Early on, it was observed
that the factor sets of Brauer [B28] and Brauer-Noether [BN] were 2-cocycles, and
the Brauer-Noether results translated immediately into the the following theorem
about the Brauer group: H 2 (G; L× ) is isomorphic to the kernel Br(L/K) of the
map Br(K) → Br(L), and is generated by the central simple algebras which are
split over L. This observation was mentioned in Eilenberg’s 1948 survey [E49] of
the field. A careful writeup was given by Serre in Cartan’s 1950/51 seminar [C50].
Shapiro’s lemma first appeared in [HN52], along with a translation of Tsen’s
Theorem (1933) into the vanishing of H q (G; K × ) for q 6= 0 when k and K are
function fields of curves over an algebraically closed field.
While studying Galois cohomology in his thesis [T52], John Tate discovered that
there is a natural isomorphism H r (G; Z) ∼
= H r+2 (G; CL ), where CL is the idèle class
group of a number field L. Moreover, the reciprocity law gave a similar relation
between H1 (G; Z) = G/[G, G] and a subgroup of H 0 (G; CL ). This led him to define
b ∗ (G, A) of any finite group G and any G-module A, indexed
the Tate cohomology H
by all integers; see [T54]. Tate did this by splicing together the cohomology of G
b r (G; A) = H r (G; A) for r > 0) and the homology of G (reindexing via Hn as
(H
b −n−1 for n ≥ 1), and using ad hoc definitions for H
b 0 and H
b −1 .
H
The 1950/51 Seminaire Cartan [C50] saw the next major advances in group
homology. In Exposés 1 and 2, Eilenberg gave an axiomatic characterization of
homology and cohomology theories for a group π, and used a fixed free resolution
of the π-module Z to establish the existence of both a homology and a cohomology
theory. The key axioms Eilenberg introduced to prove uniqueness were: 1) if A is a
free π-module then Hq (π; A) = 0 for q > 0, and 2) if A is an injective π-module then
H q (π; A) = 0 for q > 0. In Exposè 4 of the same seminar, H. Cartan proved what
we now call the Comparison Theorem for chain complexes: given a free resolution
C∗ and an acyclic resolution C∗0 of Z, there is a chain map C∗ → C∗0 over Z, unique
up to chain homotopy. This made Eilenberg’s construction natural in the choice of
C∗ , and allowed Cartan the freedom to construct cup products in group cohomology
via resolutions.
After the 1950–51 Seminaire Cartan [C50], the germs of a complete reworking
of the subject were in place. Cartan and Eilenberg began to collaborate on this
reworking, not realizing that the resulting book [CE] would take five years to appear.
HISTORY OF HOMOLOGICAL ALGEBRA
13
Associative algebras.
Before the cohomology theory of associative algebras was defined, the special
cases of derivations and extensions had been studied. Derivations and inner derivations of algebras (associative ot not) over a field k were first studied systematically
in 1937 by N. Jacobson [J37], who was especially interested in the connection to
Galois theory over k when char(k) 6= 0.
Hochschild studied derivations of associative algebras and Lie algebras in the
1942 paper [Hh42]. He showed that every derivation of an associative algebra
A is inner if and only if A is a separable algebra, meaning that not only is A
semisimple, but the `-algebra A ⊗k ` is semisimple for every extension field k ⊆ `.
In addition, he showed that if A was semisimple over a field of characteristic zero,
and f : A ⊗ A → M is a bilinear map satisfying the factor set condition:
a f (b, c) + f (a, bc) = f (a, b) c + f (ab, c).
then there was a linear map e : A → M so that f (a, b) = a e(b) + e(a)b − e(ab).
Upon seeing the Eilenberg-Mac Lane treatment of the cohomology of groups in
1945, Hochschild observed ([Hh45]) that the same formulas gave a purely algebraic
definition of the cohomology of an associative algebra A over a field, with coefficients
in a bimodule M . The degree q part C q (A; M ) of his ad hoc cochain complex is
the vector space of multilinear maps from A to M , i.e., linear maps A⊗q → M .
For example, if e : k → M has e(1) = m then δ(e)(a) = am − ma is an inner
derivation, and a 1-cocycle is a map f : A → M such that f (ab) = af (b) + f (a)b.
Thus the construction makes H 1 (A; M ) into the quotient of all derivations by inner
derivations, and the first of Hochschild’s 1942 results becomes: H 1 (A; M ) vanishes
for every M if and only if A is a separable algebra. Hochschild also showed that
H 2 (A; M ) measures algebra extensions E of A by M , meaning that M is a squarezero ideal and E/M ∼
= A; a trivial extension is one in which the algebra map
E → A splits. Since a 2-cocycle is just a map f : A ⊗ A → M satisfying the factor
set condition mentioned above, Hochschild’s second 1942 result becomes: if A is
semisimple then H 2 (A; M ) vanishes for every M , and hence every nilpotent algebra
extension of A must be split.
Lie algebras.
Since Elie Cartan’s theorem [C29] that every connected Lie group is diffeomorphic to the product of a compact Lie group G and Rn , the cohomology of Lie groups
was reduced to that of compact Lie groups. We have seen how this was solved in
1935 by Brauer and Pontrjagin. Later, Cartan and de Rham observed that the de
∗
Rham cohomology HdR
(G; R) of G may be computed using left invariant differentials, and it was gradually noticed that the Lie algebra g of left invariant vector
fields (or tangent vectors at the origin of G) determines the cohomology of G.
Chevalley and Eilenberg were able to use this observation to define the cohomology of any Lie algebra in their 1948 paper [ChE]. After reviewing de Rham
cohomology, they calculate that the (differential graded) algebra of left invariant
differential forms on a Lie group G are isomorphic to the dual algebra C ∗ (g) of the
14
CHARLES A. WEIBEL
exterior algebra ∧∗ g. Translating the de Rham differential into this context gave
the differential δ : C q (g) → C q+1 (g) defined by
(δω)(x1 , . . . , xq+1 ) =
1 X
(−1)k+l+1 ω([xk , xl ], . . . , x̂k , . . . , x̂l , . . . ).
q+1
This makes C ∗ (g) into a differential graded algebra, and they define the cohomology
∗
ring HLie
(g; R) of the Lie algebra g to be the cohomology of C ∗ (g). (They then
1
state that in other characteristics one can and should omit the constant q+1
.) Thus
if G is compact and connected then their construction of Lie algebra cohomology
∗
∗
has the isomorphism HdR
(G; R) ∼
(g; R) as its birth certificate.
= HLie
It is immediate that a 1-cocycle is a map g → R vanishing on the subalgebra
1
[g, g]. Since there are no 1-coboundaries we see that HLie
(g; k) is the dual space of
g/[g, g]. This purely algebraic feature is present, but had been downplayed in the
cohomology of compact connected Lie groups, because it follows from the fact that
G/[G, G] is a torus.
∗
In order to study the cohomology HdR
(G/H; R) of the homogeneous spaces
∗
G/H of G, Chevalley and Eilenberg also defined the cohomology HLie
(g; V ) of
a representation V of g. This was defined similiarly, as the cohomology of the
chain complex C ∗ (g; V ) of (vector space) maps from ∧∗ g to V . Translated from
the corresponding de Rham differential on the manifold G/H, the formula for the
differential δω resembled the one displayed above, but it had an extra alternating
sum of terms xk ω(· · · , x̂k , · · · ).
According to Jacobson [J37], a derivation from a Lie algebra g into a g-module
V is a linear map D : g → V such that D([x, y]) = x(Dy) − y(Dx). It is an inner
derivation if D(x) = xv for some v ∈ V . It is immediate from the Chevalley1
(g; V ) is the quotient of all derivations from
Eilenberg complex C ∗ (g; V ) that HLie
g into M by the inner derivations.
The paper [ChE] also contains the theorem that Lie extensions of g by V are in
one-to-one correspondence with elements of H 2 (g; V ), a result inspired by Eilenberg’s role in the earlier classification of group extensions via H 2 (G; A) in [EM43].
Indeed, the proof was similar: cocycles in the complex C ∗ (g; V ) are recognised as
factor sets for extensions.
Now suppose that g is any semisimple Lie algebra over a field k of characteristic zero. J. H. C. Whitehead (1904–1960) had discovered some algebraic lemmas
about linear maps on g in 1936–37 (see [W36]), in order to give a purely algebraic
proof of Weyl’s 1925 Theorem that every representation is completely reducible.
Whitehead’s lemmas also appeared in Hochschild’s paper [Hh42] on derivations.
Whitehead’s “first lemma” said that every derivation from g into any representation V was inner, even though he proved this result before the notion of derivation
was known. Chevalley and Eilenberg translated Whitehead’s “first lemma” as the
1
statement that HLie
(g; V ) = 0 for all V .
Whitehead’s “second lemma” concerned alternating bilinear maps f : g ∧ g → V
satisfying a factor set condition, which we would now write as δf (x, y, z) = 0.
Whitehead proved that for every such f there was always a linear map e : g → V
so that f (x, y) = x e(y) − y e(x) + e([x, y]). Chevalley and Eilenberg translated
2
this as the statement that HLie
(g; V ) = 0 for all V . In both of these results, the
HISTORY OF HOMOLOGICAL ALGEBRA
15
first step was an analysis of the trivial representation V = k. For the second
step, they used another result of Whitehead to show that when V 6= k is a simple
q
representation then HLie
(g; V ) = 0 for all q. This last step shows that the only
interesting cohomology groups of g are those with trivial coefficients, and these are
q
interesting because HLie
(g; k) = H q (G; k).
The analogy with the cohomology of compact Lie groups was pursued further
by Koszul (1921–) in [K50]. He introduced the notion of a reductive Lie algebra g,
and showed that (in characteristic zero) its cohomology is an exterior algebra.
Sheaves and Spectral Sequences.
Jean Leray (1906–) was a prisoner of war during World War II, from 1940 until
1945. He organized a university in his prison camp and taught a course on algebraic
topology. At the end of his imprisonment, he invented sheaves and sheaf cohomology
[L46a], as well as spectral sequences for computing his sheaf cohomology [L46b].
As we saw above, the essential features of a spectral sequence had also been
noted independently by R. Lyndon [Lyn], as a way to calculate the cohomology of a
group. The algebraic properties of spectral sequences were codified by Koszul [K47]
in 1947, using Cartan’s suggestion that the central object should be a filtered chain
complex. Koszul’s presentation clarified things so much that Leray immediately
adopted Koszul’s framework.
In 1947–48, Leray gave a course at the College de France on this new cohomology
theory. Part I was a review of his theory of spectral sequences, using Koszul’s
framework. Part II introduced the notion of a sheaf, and the cohomology of a
locally compact topological space relative to a differential graded sheaf. The details
of this course eventually appeared in Leray’s detailed article [L50].
The next year (1948-49), Henri Cartan ran a Seminar [C48] on algebraic topology,
with 17 exposés published as unbound mimeographed notes. Exposés XII–XVII
were devoted to an exposition of Leray’s theory of sheaves, but were withdrawn
when Cartan’s viewpoint on sheaves changed later that year. The same subject
was revisited by H. Cartan two years later in Exposés 14–20 of the 1950-51 Cartan
Seminar [C50], where he and his students reworked the theory of sheaves, and sheaf
cohomology, based on the notion of a “fine” sheaf.
In Exposé 16 Cartan gave axioms for sheaf cohomology theory on a paracompact
space X (with or without supports in a family Φ of closed subspaces of X, which
we shall omit from our notation here). His axioms were: H 0 (X, F ) is the group
Γ(F ) of global sections of F (with support in Φ); H q (X, F ) depends functorially on
F and vanishes for negative q; a natural long exact cohomology sequence exists for
each short exact sequence of sheaves; and if F is a “fine” sheaf then H q (X, F ) = 0
for all q 6= 0.
Cartan was now able to mimic the proof of existence and uniqueness for group
cohomology given earlier in Exposés 1–4 of the same Seminar by Cartan and Eilenberg. To prove uniqueness, he observed that every sheaf F may be embedded in
a fine sheaf, specifically into a sheaf he called F ⊗ S, which we would describe
as the sum of the skyscraper sheaves x∗ x∗ (F ) over all points x of X. To prove
existence, Cartan fixed a resolution 0 → Z → C0 → · · · of Z by torsionfree fine
sheaves, and set H q (X, F ) = H q (Γ(C ⊗F )). Observing that some choices of C happen to give differential graded algebras, he was able to define a product structure
16
CHARLES A. WEIBEL
H p (X, F ) ⊗ H q (X, F 0 ) → H p+q (X, F ⊗ F 0 ) on sheaf cohomology.
In the remaining exposés of [C50], Cartan, Eilenberg and Serre returned to
Leray’s spectral sequences, codifying the machinery and studying its multiplicative
structure. Much of this material was reproduced in the Hochschild-Serre paper
[HS53] in order to redo Lyndon’s spectral sequence [Lyn]. The usefulness of this
approach to spectral sequences was decisively demonstrated by Serre in his 1951
thesis [Se51].
A completely different approach to spectral sequences was given by W. Massey
in 1952 ([M52]). Massey defines an exact couple to be a pair of (graded) modules
D and E, equipped with maps fitting into an exact sequence
i
j
k
i
D−
→D−
→E−
→D−
→ D.
One forms its derived couple by considering D1 = i(D) and the homology E1 of E
with respect to the differential jk. By an iterative process, one obtains a sequence
of derived couples, and the sequence of modules Er forms a spectral sequence. The
exact couple approach to spectral sequences has since become very popular with
topologists, but less so with algebraists.
Godement’s 1958 book [Gode] summarized and refined all these developments,
becoming the standard reference for sheaves, sheaf cohomology and spectral sequences for many years. In Godement’s approach, the focus moved away from
Cartan’s notion of “fine” sheaf and towards the new notions of flasque and soft
(mou) sheaves. One trick introduced by Godement, but implicit in Cartan’s 1950–
51 seminar [C50], was that by iteration of the canonical embedding F of into F ⊗ S
one could get a resolution of F by injective sheaves which is functorial in F ; nowadays it is called the Godement resolution of F .
The Cartan–Eilenberg Revolution
As we have mentioned, Cartan and Eilenberg began collaborating during the
1950–1951 Seminaire Cartan [C50], rewriting the foundations of all the ad hoc
algebraic homology and cohomology theories that had arisen in the previous decade.
Coining the term Homological Algebra for this newly unified subject, and using it
for the title of the textbook [CE], they revolutionized the subject.
The first occurrence of the notation Torn and Extn , as well as the concepts of
projective module, derived functor and hyperhomology appeared in this book. In
his review of their book, Hochschild stated that “The appearance of this book must
mean that the experimental phase of homological algebra is now surpassed.”
Before we describe the innovations in their book further, let us back up and
review the evolution of the two main tools that were now available, namely chain
complexes and resolutions.
Chain Complexes.
The algebra of chain complexes had been slowly evolving since their formal
introduction in 1929 by Mayer [M29]. We have already mentioned Hurewicz’ 1941
discovery of the notion of exact sequence ([H41]), and the application of this notion
in the 1945 axiomatization of homology theory [ES45].
HISTORY OF HOMOLOGICAL ALGEBRA
17
The next step was taken in 1947 by Kelley and Pitcher [KP], who coined the term
“exact sequence” and first systematically studied chain complexes from an algebraic
point of view. They showed that direct limits preserve exact sequences (axiom AB5
holds), but that inverse limits do not (axiom AB5* fails). If A∗ is a subcomplex of
B∗ , with quotient C∗ , they constructed the boundary map ∂ : Hq (C) → Hq−1 (A)
and proved that the long homology sequence
∂
· · · → Hn (A) → Hn (B) → Hn (C) −
→ Hn−1 (A) → · · ·
is exact. Since they restricted themselves to positive complexes (indexed by positive
integers q), their sequence ended in H0 (B) → H0 (C) → 0.
The yoga of chain complexes was further developed in Eilenberg and Steenrod’s
1945 book [ES]. They indexed their chain complexes by all integers, and observed
that cochain complexes could be identified as chain complexes via the reindexing
Cq = C −q . The familiar “five-lemma” occurs for the first time on p. 16 of [ES].
(Its companion, the “snake lemma,” first appeared in [CE].) Eilenberg and Steenrod’s book also introduced the “Mayer-Vietoris” sequence for a space X = U ∪ V ,
associated to the excision isomorphism H∗ (U, U ∩ V ) ∼
= H∗ (X, V ).
Free and Injective resolutions.
Free resolutions have long been used in algebra, starting with David Hilbert
(1862–1943) in his 1890 paper [Hilb] on iterated syzygies of a finitely generated
graded module M over a polynomial ring R = k[x1 , . . . , xn ]. A choice of b0 =
dim(M ⊗R k) homogeneous generators of M defines a surjection R b0 → M , and its
kernel is the first syzygy module of M . (There is a grading on R b0 which we are
ignoring.) Hilbert proved that the syzygy was also finitely generated (the Hilbert
Basis Theorem), so one could use induction to define the higher syzygy modules.
Hilbert’s Syzygy Theorem states that the (n + 1)st syzygy is always zero, i.e., the
nth syzygy is Rbn for some bn . Since the number of generators bi of the syzygies
is chosen minimally, they are independent of the choices of generators: today we
know this is so because bi is the dimension of the vector space TorR
i (M, k). By
analogy with topology, the bi are called the Betti numbers of M .
As we have remarked, Baer [B34] implicitly used free resolutions of an abelian
group to study the groups Ext(A, B). The next explicit use of free resolutions was
by Hopf in 1944 [Hf44]. As we have mentioned above, he used them to describe the
homology of a group, and implicitly gave a definition of the modules TorR
i (M, R/I)
for any ideal I of any ring R. Based on Hopf’s work, Cartan and Eilenberg used
free Z[π]-resolutions of a π-module A in [C50] to give an axiomatic description for
the group homology H∗ (G; A).
Injective R-modules were introduced and studied in 1940 by R. Baer [B40].
Baer called them “complete” abelian groups over the ring R; the name injective
apparently first arose in Eilenberg’s survey paper [C50]. Baer’s paper contains the
proposition that every module is a submodule of an injective module, and what is
now called “Baer’s criterion” for M to be injective: every map from an ideal into
M must extend to a map from R into M . Finally, Baer characterized semisimple
rings as those for which every module is injective.
In the 1948 paper [M48], Mac Lane formulated the projective and injective lifting
properties for the category of abelian groups, and showed that these properties
18
CHARLES A. WEIBEL
describe free and divisible abelian groups, respectively. He did not discover the
notion of projective module because he did not apply these lifting properties to
categories of modules. Using this, he showed that one could compute Ext(A, B) by
embedding the abelian group B in a divisible group D; this amounts to the use of
an injective resolution of B.
Cartan and Eilenberg: the book.
We now turn to the contents of the book [CE] itself. On p. 6 we find an entirely new concept: the definition of a projective module. By p. 11 we find that
every R-module is projective if and only if R is semisimple, complementing Baer’s
characterization of semisimplicity in terms of injective modules; later in the book
(p. 111), this is viewed as the characterization of rings of global dimension 0.
In chapter II the authors introduce the notion of left exact functors (such as Hom)
and right exact functors (such as ⊗R ). In the central chapter V, they introduce the
notions of projective resolutions · · · → P0 → M and injective resolutions M → I 0 →
· · · of a module M , and use these to define the derived functors Ln T (M ) = Hn T (P∗ )
and Rn T (M ) = H n T (I ∗ ) of an additive functor T . This material is clearly based
on the ideas in the 1950–1951 Seminaire Cartan [C50].
n
In chapter VI, the authors define TorR
n (M, N ) and ExtR (M, N ) as the derived
functors of M ⊗R N and HomR (M, N ). Then they define the projective and injective dimension of M as the length of the shortest projective and injective resolution, and characterize these dimensions in terms of the vanishing of ExtnR (M, −) and
ExtnR (−, M ), respectively. This led them to define the (left and right) global dimension of R as the largest n such that ExtnR is nonzero, and the weak global dimension
(now called the Tor-dimension) as the largest n such that TorR
n is nonzero.
Chapters VIII to XIII unified the homology of augmented algebras, Hochschild’s
homology and cohomology of associative algebra Λ (as Tor and Ext groups over the
enveloping algebra Λ ⊗ Λop ), the homology and cohomology of a group π (as Tor
and Ext groups over the group ring Z[π]), and the homology and cohomology of a
Lie algebra g (as Tor and Ext groups over the enveloping algebra U g).
Chapters XV–XVI contained a very readable introduction to spectral sequences
for filtered chain complexes, and applications to computing Ext and Tor. Again,
this material is based on the ideas in the 1950–1951 Seminaire Cartan [C50].
The final Chapter (XVII) concerned the hyperhomology of a functor T applied
to a chain complex A. This was the precursor to the discovery of the Derived
Category by Grothendieck and Verdier [V]. First they defined double complexes
they called “projective” and “injective” resolutions of A; since [HRD] we call them
Cartan-Eilenberg resolutions of A. Then they defined the hyperhomology L ∗ T (A)
and hypercohomology R∗ T (A) to be the (co)homology of the total complex of T
applied to the double complex resolutions.
Until 1970, [CE] was the bible on homological algebra, although Mac Lane’s
book [ML] was also popular. Grothendieck’s Tohoku paper [G57], which we shall
describe below, and later his multi-volume tome [EGA] on the foundation of sheaf
cohomology in Algebraic Geometry, were also heavy favorites. In 1970, Rotman’s
Notes on Homological Algebra appeared, and Hilton and Stammbach’s book [HStm]
appeared in 1971. At that point, the subject was firmly established.
HISTORY OF HOMOLOGICAL ALGEBRA
19
Abelian Categories.
As soon as Cartan and Eilenberg began their undertaking, limiting theselves to
functors defined on modules, it was clear that there was more than a formal analogy
with the cohomology of sheaves, and that their methods worked in a more general
setting. The search for that setting led to the notion of an abelian category.
The first attempt to formulate a setting in which homological algebra made
sense was by Mac Lane in 1948 [M48]. In this paper Mac Lane introduced what he
called “abelian categories,” but which were actually additive categories with special
objects resembling the objects Z and Q/Z in the category Ab of abelian groups.
The category of abelian semigroups was an abelian category in Mac Lane’s sense,
and his notion never caught on.
The appendix to [CE] contained the next attempt, by D. Buchsbaum. It was actually a summary without proofs of his 1955 thesis [B55], written under Eilenberg.
In attempting to formulate a general setting in which the theory in Cartan-Eilenberg
could be generalized, he needed categories which had a natural notion of an exact
sequence. To this end, Buchsbaum introduced the notion of an exact category,
which is an abelian category without the requirement that direct sums exist. To
handle functors of more than one variable, he introduced the extra axiom (V) that
direct sums A ⊕ B exist, which is equivalent to the definition of an abelian category. Buchsbaum also introduced axioms that the category has enough projectives
or enough injectives. These axioms, unnecessary for the categories of modules considered in [CE], allowed Buchsbaum to carry over verbatim the Cartan-Eilenberg
construction of derived functors to exact categories.
The name abelian category is due to A. Grothendieck [G57] and A. Heller [H58].
Grothendieck’s paper was motivated by the observation that the category Sh(X)
of sheaves of abelian groups on a topological space X was an abelian category
with enough injectives, so that sheaf cohomology could be defined as the right
derived functors of the global sections functor, while Heller was more concerned
with a formal analogy to stable homotopy (where syzygy modules correspond to
loop spaces, and projective modules correspond to contractible spaces).
Grothendieck’s 1957 “Tohoku” paper [G57] introduced a hierarchy of axioms
(AB3)–(AB6) and (AB3*)–(AB6*) that an abelian category may or may not satisfy. Axioms (AB3) and (AB3*) specify that set-indexed coproducts and products
exist, respectively. The abelian category Sh(X) satisfies axiom (AB5), that filtered
colimits of exact sequences are exact, but not axiom (AB4*), which states that a
product of surjections is a surjection.
Given this framework, Grothendieck proceeded to generalize Cartan and Eilenberg’s treatment of derived functors, introducing the names ∂-functor and universal
∂-functor, as well as the notion of T -acyclic objects (in [CE, p. 122] flat modules
were defined as Tor-acyclic modules; Grothendieck showed that Godement’s flasque
sheaves were Γ-acyclic sheaves). The primary computational tool introduced by
Grothendieck was a special case of the hypercohomology spectral sequence for the
composition T U of two functors (see the last page of [CE]). Grothendieck observed
that if T and U were left exact, and if U sends injective modules to T -acyclic
modules then we could write the spectral sequence as
(Rp T )(Rq U ) =⇒ Rp+q (T U ).
20
CHARLES A. WEIBEL
Several of the spectral sequences in [CE] were seen to be simple special cases of
Grothendieck’s spectral sequence, but so were the Leray spectral sequences associated to a continuous map f : Y → X and a sheaf F on Y :
H p (X, Rq f∗ F ) =⇒ H p+q (Y, F ).
Even the simplest of lemmas (such as the Snake Lemma) were painfully difficult
to prove in a general abelian category, because one couldn’t chase elements that
didn’t exist. This technique of diagram-chasing was justified in 1960, when Saul
Lubkin [L60], A. P. Heron (1960 Oxford thesis) and J. P. Freyd (1960 Princeton
thesis) proved that every small abelian category admits an exact embedding into the
category of abelian groups. Shortly thereafter, Freyd and Barry Mitchell proved
a stronger version: every small abelian category admits a full exact embedding
into the category of modules over some ring (see [F64]). With this result, and
P. Gabriel’s thesis [G62], the subject was near maturity.
After the Cartan–Eilenberg Revolution
Upon the publication of Cartan-Eilenberg [CE], there was an explosion of research in homological algebra. Some results appeared to be fairly isolated curiousities at the time, but later became important, such as Yoneda’s definition of
Extn groups by long exact sequences in [Y54], the 1961 study of lim1 by J. E. Roos
[R61], the Eilenberg-Moore paper [EM62] on spectral sequences for complete filtered complexes, Giraud’s work [G65] on nonabelian H 1 in a Grothendieck Topos,
or Boardman’s influential preprint [B81] on conditional convergence in spectral sequences. In this article we shall focus upon the strands of thought that have led to
flourishing new fields of study.
Projective Modules.
When the notion of projective module was introduced in [CE], there were not
many examples of projective modules which were not free. By [CE, p. 157], all
finitely generated projective modules over a local ring are free. By [CE, p. 13],
all projective modules over a principal ideal domain (or more generally a Bezóut
domain) are free. Kaplansky later showed [K58] that all projective modules over a
local ring are free, as a consequence of the general result that any infinitely generated projective module is a direct sum of countably generated projective modules.
If I is an ideal of an integral domain R, Cartan and Eilenberg showed that I was
projective if and only if it was invertible: I · I −1 = R. Moreover, if dim(R) = 1
then invertible ideals have two generators, so I ⊕ I −1 ∼
= R ⊕ R. Since every ideal
in a Dedekind domain is invertible — their isomorphism classes forming the Picard
class group of R — and the integers in a number ring were Dedekind domains whose
class groups were classical objects of study, some examples of non-free projective
modules were already known in the late 19th century.
For some rings, it was possible to classify all projective modules. A Prüfer
domain is a commutative domain in which every finitely generated ideal is invertible;
this generalization of Dedekind domains is named for H. Prüfer, who initiated their
study in 1923. Kaplansky [K52] showed that if R is a Prüfer domain then every
finitely generated torsionfree module — hence every projective module — is a direct
sum of invertible ideals; see [CE, pp. 13, 133].
HISTORY OF HOMOLOGICAL ALGEBRA
21
For other rings, the classification was much harder. In Serre’s classic 1955 paper
[Se55, p.243], he stated that it was unknown whether or not every projective Rmodule was free when R is a polynomial ring over a field. This became known
as the “Serre problem,” and was not solved (affirmatively) until 1976, by Quillen
[Q76] and Suslin [S76].
In the period 1958–1962 there was a flurry of examples of non-free projective
modules, coming from algebraic geometry [BS, Se58], arithmetic [B61], group rings
[Sw59] and topological vector bundles [Sw62]. Much of this was based upon the
dictionary in Serre’s 1955 paper [Se55], between projective modules and topological
vector bundles. Grothendieck’s Riemann-Roch Theorem, published in [BS], showed
that the “projective class group” K(R) of stable isomorphism classes of projective
modules was useful, especially for rings coming from algebra and algebraic geometry.
Bass, Serre and Swan began a study of the projective class group K(R); by 1964
it was renamed K0 (R) in view of its parallels to topological K-theory, and this led
to the rise of algebraic K-theory in the 1960’s.
Homological Algebra and ring theory.
The left and right global dimension of a ring were early targets. In [A55], M. Auslander (1926–1994) showed that the left and right global dimension of a noetherian
ring agree, and equal the weak global dimension. Then M. Harada [H56] showed
that the rings with weak global dimension 0 are precisely the von Neumann regular
rings, so the weak dimension and global dimension need not agree. Examples in
which the left and right global dimensions of a ring are different were not known
until a decade later, and were found by Osofsky [O68].
Regular local rings.
A regular local ring is a commutative noetherian local ring R whose maximal
ideal m is generated by a regular sequence, or equivalently, such that dim(m/m 2 ) =
dim(R). Regular local rings had become important in algebraic geometry because
they were the local coordinate rings of smooth algebraic varieties. Auslander and
Buchsbaum [AB56] and Serre [Se56] used homological methods to characterize regular local rings as those (noetherian) local rings R with finite global dimension. If R
is local with residue field k and dimk (m/m2 ) = n, Serre proved that TorR
n (k, k) 6= 0.
Hence gl. dim(R) ≥ n, and n ≥ dim(R) ≥ depth(R). Auslander and Buchsbaum
proved that the depth of R is an upper bound for the finite values of pdR (M ), so
if pdR is always finite we must have equality: gl. dim(R) = dimk (m/m2 ) = dim(R).
In particular, if gl. dim(R) < ∞ then R must be regular.
Since localization cannot increase global dimension, a corollary is that any localization of a regular local ring is again a regular ring. This non-homological
statement, proven by homological methods, firmly established homological algebra as a central tool in ring theory; the alternate non-homological proof of this
localization result, due to Nagata [N58], is very long and hard.
Also in [AB56], Auslander and Buchsbaum proved that 2-dimensional regular
local rings are Unique Factorization Domains (UFD’s). A few years later, Auslander and Buchsbaum [AB59] used similar homological methods to prove that every
regular local ring is a Unique Factorization Domain.
22
CHARLES A. WEIBEL
Two timely courses on this material, by Serre in France and Kaplansky in the
U.S., had a lasting impact upon the field.
In 1957–58, Serre taught a course on multiplicities at the Collège de France [SeM].
Part of that course focussed upon the simple inequality pdR (M ) ≤ pdR (S)+pdS (M )
for a module M over an R-algebra S (an exercise on p. 360 of [CE].) Auslander and
Buchsbaum realized that Serre’s methods could be used to study the connection
between the codimension and multiplicity over a local ring; see [AB58]. This led
them to the Auslander-Buchsbaum Equality: if M is a finitely generated module
over a local ring R and pdR (M ) < ∞ then depth(R) = depth(M ) + pdR (M ).
In Fall 1958, Kaplansky taught a course [K59] on homological algebra at the
University of Chicago. Several students attending this course would later make
important contributions to the subject: H. Bass, S. Chase, E. Matlis and S. Shanuel.
Kaplansky’s course was organized around three “change of rings” theorems, describing how homological dimension changes when one passes from a ring R to
a quotient ring R/(x). They allowed him to prove the Theorems of Serre and
Auslander-Buchsbaum without having to first develop Ext or Tor. Early in the
course, Shanuel noticed that there was an elegant relation between different projective resolutions of the same module. Kaplansky seized upon this result as a way to
define projective dimension, and christened it “Shanuel’s Lemma.” Subsequently
it was discovered that H. Fitting had proven Shanuel’s Lemma in 1936 [F36] (with
“projective” replaced by “free”) as part of his study of the Fitting Invariants of a
module.
Tor∗ (k, k) for local rings.
Consider a local ring R with maximal ideal m and residue field k = R/m.
Cartan and Eilenberg had shown that TorR
∗ (k, k) was a graded-commutative kalgebra [CE, XI.4–5]. Its Hilbert function is just the sequence of Betti numbers bi =P
dim TorR
i (k, k), and it is natural to consider the Poincaré-Betti series
∞
i
2
PR (t) =
b
t
i . Note that the first Betti number is b1 = dim(m/m ). For
i=0
example, if R is a regular ring, it was well known that TorR
∗ (k, k) was an exterior
b1
algebra, so that PR (t) = (1 + t) .
Serre showed
in 1955 [Se56] that one always had PR (t) ≥ (1 + t)b1 , i.e., that bi
is at least bi1 . In particular, if i = b1 then bi ≥ 1 and so TorR
b1 (k, k) 6= 0. As we
mentioned above, this was the key step in Serre’s proof that local rings of finite
global dimension are regular. In his 1956 study [T57], Tate showed that k had a
free R-module resolution F∗ which was a graded-commutative differential graded
algebra, and used this to show
that if R is not regular then PR (t) ≥ (1+t)b1 /(1−t2 ),
b1
b1
+ · · · . This is the best lower bound. In case R
i.e., that bi is at least i + i−2
is the quotient of a regular local ring by a regular sequence of length r (contained
in the square of the maximal ideal), Tate showed that the Poincaré-Betti series of
R is the rational function PR (t) = (1 + t)b1 /(1 − t2 )r .
Based upon Tate’s results, Serre stated on p. 118 of [SeM] that it was not
known whether or not PR (t) was always a rational function. This problem remained open for over twenty years, until it was settled negatively by David Anick
[A82]. Anick’s example was an artinian algebra R with m3 = 0. Constructing
a finite simply-connected CW complex X whose cohomology ring was R, a result of Roos [R79] showed that the Poincaré-Betti series of the loop space ΩX,
HISTORY OF HOMOLOGICAL ALGEBRA
23
P
H(t) =
dim Hi (ΩX) ti was not a rational function either. This settled a second
problem of Serre, also posed on p. 118 of [SeM].
Matlis Duality.
In his 1958 thesis [M58] under Kaplansky, Eben Matlis studied the structure of
injective modules over a noetherian ring R, and showed that they can be written
uniquely as direct sums of copies of the injective hulls E(R/p), as p ranges over the
prime ideals of R. This put injective resolutions on an equal footing with projective
resolutions.
Let A denote an additive category of modules over a ring R. A dualizing functor on A is an exact contravariant R-linear functor D from A to itself such that
D(D(M )) = M . Matlis’ thesis [M58] also showed that the category A of modules of finite length over a local noetherian ring R has a unique dualizing functor:
D(M ) = HomR (M, E), where E is the injective hull of R/m.
This turned attention to other kinds of duality, and to modules of finite injective
dimension. The goal here was to find the analogue of Serre’s Duality Theorem for
projective space X = Pd [Se55]: if F is a coherent sheaf on X then the dual of the
d−i
vector space H i (X; F ) is ExtX
(F, ωX ), where ωX = ΩdX is the sheaf of differential
d-forms on X.
It would turn out that that the good class of rings from this perspective would be
Gorenstein rings. In a 1957 Séminaire Bourbaki talk on Duality ([GFGA, exp. 2],
Grothendieck defined a commutative ring R (or scheme) of finite type over a field to
be “Gorenstein” if it is Cohen-Macaulay and a certain R-module ωR is locally free
of rank 1. A few years later, Bass proved a theorem characterizing rings of finite
self-injective dimension, and Serre remarked that the two definitions agreed in a
geometric context. Bass then consolidated these notions in [B62], giving the modern
definition: a commutative noetherian ring R is called Gorenstein if all its local rings
have finite injective dimension. Bass proved that this is equivalent to several other
conditions, such as R being Cohen-Macaulay and a system of parameters generates
an irreducible ideal in each local ring. Nowadays we have the notion of the canonical
module ωR of a ring (see below), and if R is a Cohen-Macaulay local ring, then R
is Gorenstein if and only if R is its own canonical module: ωR = R. For example,
in Matlis Duality for a zero-dimensional ring, the role of ωR is played by E, and R
is Gorenstein exactly when E = R.
Local Cohomology and Duality.
In 1961, Grothendieck ran a Harvard seminar on Local Cohomology, based upon
his 1957 Séminaire Bourbaki talk on Duality ([GFGA, exp. 2]); the notes were eventually published in [G67]. ¿From the viewpoint of schemes, the local cohomology
of a sheaf is the same as cohomology with supports. From the viewpoint of noe∗
therian local rings, the local cohomology Hm
(M ) of a module M are the derived
0
functors of the m-primary submodule functor Hm
(M ) = lim
HomR (R/mn , M ), so
−
→
i
i
n
Hm
(M ) = lim
−→ ExtR (R/m , M ).
Grothendieck showed that the depth of M is characterized as the smallest i such
i
i
that Hm
(M ) 6= 0, and that if R is a Cohen-Macaulay ring then Hm
(R) 6= 0 only for
dim(R)
i = dim(R). Moreover, R is a Gorenstein ring if and only if the module Hm
(R)
is dualizing in Matlis’ sense, meaning that it is the injective hull of R/m.
24
CHARLES A. WEIBEL
The highlight of the seminar was the Duality Theorem: if R is a complete Gorend−i
i
stein ring of dimension d, then Hm
(M ) is dual to ExtR
(M, R), in the sense that
Matlis’ dualizing functor D interchanges them. For a more general local ring, the
duality is more complicated. If R is complete and Cohen-Macauley, one considi
ers the functors T i (M ) = D(Hm
(M )), and shows that they equal Extd−i
R (M, ωR ),
d
where ωR = D(Hm (R)). More generally, Grothendieck also observed that the
d−i
T i (M ) may be interpreted as ExtR
(M, KR ) for a suitable dualizing cochain complex KR on R [G67, 6.8]. This led to the development of the derived category
D(R), which we shall describe shortly.
This material on Duality took awhile to absorb, and a ring-theoretic derivation
of these results only appeared in 1970 [S70]. Gradually the notion of a canonical
module ωR became the organizing principal for duality theory, and R is Gorenstein
exactly when ωR = R. If R is Cohen-Macaulay, the canonical module is defined
[HK71] to be a maximal Cohen-Macaulay R-module of finite injective dimension,
and the functor D(M ) = HomR (M, ωR ) is dualizing on the category of maximal
Cohen-Macaulay R-modules.
In 1971, Sharp [S71] used local cohomology (and duality) to show that if R is a
complete Cohen-Macauley local ring then the Gorenstein modules are precisely the
direct sums of ωR . He also showed that the final term in the Cousin complex of an
dim(M )
(M ).
R-module M is Hm
In 1976 Hochster and Roberts [HR76] studied the local cohomology of a graded
ring R in characteristic p > 0, and found that the structure of the local cohomology
i
Hm
(R) was amazingly simplified under certain assumptions, such as the purity
of the Frobenious homomorphism F : R → R. They were also able to lift these
characteristic p results to certain rings of characteristic 0, beginning a rennaisance
in the study of Cohen-Macaulay rings.
Cohomology Theories in Algebraic Geometry.
During the early 1950’s, the foundations of algebraic geometry were reworked
by O. Zariski and others, focussing upon the role played by the algebras of regular
functions. In his classic paper [Se55], Serre observed that if U is affine, with coordinate ring R, then there is an equivalence between finitely generated R-modules
and coherent sheaves of modules on U . Hence restriction to an affine open V of U
is an exact functor on coherent modules, because it corresponds to localization of
modules. This implies that if F is coherent and U is affine then the Čech cohomology Ȟ q (U, F ) vanishes. Using this, Serre defined the cohomology groups H q (X, F )
of a coherent module on any variety X as the Čech cohomology relative to a covering of X by affine open subvarieties U . All this was in the spirit of the Cartan
Seminars on Sheaf Theory in 1948–1950, but with the homological underpinnings
of Cartan-Eilenberg available, Serre’s presentation in terms of the Zariski topology
was much simpler.
Serre also proved in [GAGA] that if X is a projective variety over C the groups
H q (X, F ) were the same as the analytically defined Betti cohomology, leaving little
doubt that using the Zariski topology was a good approach to cohomology.
Grothendieck then observed that Serre’s construction was a special case of the
derived functor sheaf cohomology (for the Zariski topology) that he had developed
HISTORY OF HOMOLOGICAL ALGEBRA
25
in [G57]. Chapter III of [EGA] was devoted to the Zariski cohomology theory of
coherent sheaves on a scheme, using the right derived functors Rf∗ associated to a
morphism f : X → Y .
As part of the preliminaries to this development, Grothendieck wrote a primer
on spectral sequences and hypercohomology in [EGA, 0III ]. This was a reworking
of the corresponding material in [CE] and [G57] into a more workable form, and
made these tools widely available to algebraic geometers.
Galois cohomology.
We have already mentioned that Hochschild [Hh50] coined the term “Galois
cohomology” for the group cohomology of the Galois groups G = Gal(K/k), where
K is a (possibly infinite) Galois field extension of k. As we have already mentioned,
Hochschild and Tate ([T52], [T54]) applied Galois cohomology to class field theory.
In the 1950’s Tate began to systematically study what he called the “Galois
cohomology” of the Galois groups G = Gal(K/k), where K is a (possibly infinite)
Galois field extension of k, such as the separable closure of k. Such a group has a
topology induced by its finite quotients: G = lim
←− G/HF , where F ranges over all
the finite extensions of K contained in k and HF = Gal(K/F ). As a topological
group, G is compact, Hausdorff and totally disconnected; today we call such groups
profinite. Moreover, each HF is an open subgroup of finite index in G.
In 1954, Kawada and Tate [KT55] used Galois cohomology to calculate the
cohomology of a variety. To an étale covering U of X they associated a subgroup
of the Galois group of k(U )/k(X). This would later be recognized as the first use
of what would later be called étale cohomology.
After years of gestation, a published account of Galois cohomology appeared in
the 1958 paper [LT] by Serge Lang and John Tate. One considers a G-module
A which is discrete in the sense that the action G × A → A is continuous (when
A has the discrete topology), and defines the Galois cohomology H ∗ (G, A) to be
the cohomology of the complex C ∗ (G, A) of continuous cochains, that is, maps
φ: Gn → A which are continuous. An almost immediate observation is that
∗
H
H ∗ (G, A) = lim
−→ H (G/H, A )
H
as H ranges through the open subgroups of finite index in G.
Tate’s applications lay in the cases where A is an abelian group scheme defined
over k; the G-module in this case is A = A(k̄), the group of rational points over
the separable closure k̄ of k.
One of the most important examples is the group scheme A = Gm , for which
the G-module A is k̄ × = Gm (k̄) of units of k̄. Hilbert’s “Theorem 90” states that
for every finite Galois extension F/k we have H 1 (Gal(F/k), F × ) = 0; taking the
direct limit over all such F and setting G = Gal(k̄/k) yields the infinite version
H 1 (G, k̄ × ) = 0. As we have seen, it was already known that H 2 (Gal(F/k), F × ) is
the relative Brauer group Br(F/k); taking the direct limit over all such F shows
that H 2 (F, Gm ) is the classical Brauer group Br(F ) introduced by Richard Brauer
[B28] and Brauer-Noether [BN].
Serre’s 1962 course Cohomologie galoisienne [SeCG], published in 1964, has remained the standard reference on the Galois cohomology over number fields.
26
CHARLES A. WEIBEL
Étale cohomology.
In 1958, Grothendieck found a common generalization of Galois cohomology
and Zariski cohomology and used it to define the étale cohomology of schemes. A
Grothendieck topology is a category T such that each object X is equipped with a
family of morphisms {Ui → X}, called coverings, subject to certain axioms. From
this viewpoint, a sheaf F is a contravariant functor on T such that for each covering,
each s ∈ F (X) is uniquely determined by elements si ∈ F (Ui ) which agree in each
F (Ui ×X Uj ). The category of sheaves of abelian groups on T is an abelian category
with enough injectives, and Grothendieck defined the cohomology groups H ∗ (T , F )
to be the right derived functors of F 7→ F (X). When X is a topological space and
T is the poset of open subspaces then sheaf has its usual meaning, and we recover
the usual sheaf cohomology on X.
To define the étale topology on a scheme X, Grothendieck took the category of all
schemes U which are étale over X, with the set-theoretic notion of covering. If F is a
sheaf for this topology, the above construction defines the étale cohomology groups
H ∗ (Xet , F ) of F on X. When X is the spectrum of a field k and G = Gal(k̄/k), a
discrete G-module A is the same as an étale sheaf on X, so the étale cohomology
of X with coefficients A agrees with Tate’s Galois cohomology H ∗ (G, A).
In Fall 1961, Grothendieck presented his ideas in a course at Harvard. The
following semester (Spring 1962), M. Artin ran a seminar covering Grothendieck
Topologies, as well as some material on étale cohomology (such as cohomological
dimension). The published notes [A62] of this seminar, as well as Giraud’s Bourbaki
talk [Gir63] made the ideas available to a wide audience.
The next year (1962–63), when the seminar continued in France, Artin and
Grothendieck worked out the fundamental structure theorems of étale cohomology:
proper and smooth base change, specialization, cohomology with compact supports
and duality. The following year, more results were obtained (such as purity and the
Lefschetz trace formula), with the seminar notes eventually appearing as [SGA4].
One of Grothendieck’s early successes with étale cohomology was his cohomological proof of the rationality of the Zeta function ZX (t) of a scheme of finite type
over the finite field Fq . He proved that each factor Pi (t) of ZX (t) is the characteristic polynomial of the Frobenius operator acting on an l-adic cohomology group,
i
ν
namely H i (X, Ql ) = lim
−→ Het (X, Z/(l )). In 1972, Deligne used étale cohomology
to prove the “Riemann hypothesis” over Fq [D74]: the eigenvalues of the Frobenius
on H i (X, Ql ) (and hence the zeroes and poles of the zeta function) were algebraic
integers with absolute value q i/2 . This completed the proof of the celebrated Weil
Conjectures, and firmly established the importance of étale cohomology.
Derived Categories.
After Grothendieck’s 1961 Harvard seminar on Local Cohomology, described
above, Grothendieck realized that in order to extend these results to arbitrary
schemes he needed some results in homological algebra which were not yet available.
This was overcome by Verdier’s 1963 thesis [V] on Derived Categories.
The derived category D(A) of an abelian category A is the category obtained
from the category Ch(A) of (co)chain complexes by formally inverting the quasiisomorphisms, i.e., the maps C → C 0 which induce isomorphisms on (co)homology.
HISTORY OF HOMOLOGICAL ALGEBRA
27
To describe it, Verdier introduced the notion of a triangulated category. The quotient category of Ch(A) whose morphisms are the chain homotopy equivalence
classes of maps is triangulated; D(A), which is formed from this by a calculus of
fractions, is also triangulated. If F : A → B is an additive functor then under reasonable conditions there is a functor RF : D(A) → D(B) with the property that if
an A in A is considered as a complex then the cohomology of the complex RF (A)
give the ordinary right derived functors R∗ F (A).
In the Summer of 1963, after Hartshorne proposed to run a seminar at Harvard on duality theory, Grothendieck wrote a series of “prenotes,” sketching the
construction of a functor f ! : D(Y -mod) → D(X-mod) associated to a reasonable morphism f : X → Y of schemes, together with a natural trace morphism
Rf∗ f ! (A) → A. The so-called “Séminaire Hartshorne” was held at Harvard in
1963–64, based upon these prenotes, and the seminar notes appeared as [H66]. An
appendix to [H66], written by Deligne in 1966, constructs f ! for every separated
morphism of finite type between noetherian schemes.
During the 1966–67 Séminaire de Geometrie Algebrique [SGA6], Grothendieck
used the triangulated category Perf(X) of perfect complexes of OX -modules to
develop a global theory of intersections and a Riemann-Roch Theorem for arbitrary noetherian schemes. By definition, a complex is perfect if it is locally quasiisomorphic to a bounded complex of vector bundles, and the alternating sum of
these vector bundles gives a well-defined element in the Grothendieck group K(X),
at least if X is quasi-projective or smooth. If f : X → Y is proper, there machinery of triangulated categories yields an exact functor Rf∗ : Perf(X) → Perf(Y ) and
hence a homomorphism K(X) → K(Y ).
In 1978, Bernstein-Gelfand-Gelfand [BBG] used derived categories to classify
vector bundles on projective space Pn over a field k in terms of graded modules
over the exterior algebra Λ on n+1 variables. The crucial step in their classification
was the discovery of an isomorphism between the (bounded) derived categories of
b
b
graded modules Dgr
(Λ) and Dgr
(R), where R is the polynomial algebra on n + 1
variables. This result showed that D b (A) did not determine the “heart” category
A, a result which came as a bit of a surprise.
The problem of multiple hearts for a triangulated category was revisited in 1982
by Bernstein-Beilinson-Deligne [BBD]. These authors used triangulated categories
to study D-modules and perverse sheaves on a stratified space. In 1988, BeilinsonGinsburg-Schechtman [BGS] generalized the results of [BBG] and [BBD] by proving
that many filtered triangulated categories have two hearts, which are in Koszul
duality.
In the mid-1980’s, derived categories found yet another application. The notion
of a tilting module had come up in the study of representations of finite algebras.
Cline-Parshall-Scott [CPS] showed that if T is a tilting module for A, and B =
HomA (T, T ), then D b (A) ∼
= D b (B).
Early work on derived categories was often restricted to either bounded or
bounded below complexes, because of the need to work with injective (or projective)
resolutions. In 1988, Spaltenstein [S88] showed that every unbounded complexes
was quasi-isomorphic to a “fibrant” complex, and that one could use fibrant complexes to compute derived functors. This result has led to several new developments
which continue to this day.
28
CHARLES A. WEIBEL
Simplicial Methods
During the 1940’s, Eilenberg kept encountering things called “abstract complexes” which resembled the triangulated polyhedra (or “geometric simplicial complexes”) introduced by Poincaré, except that a simplex was not always determined
by its faces. For example the abstract complex K(π) of [EM43] and the singular
complex S(X) of [E44] had this property. To describe this phenomenon, Eilenber
and Zilber [EZ50] introduced the notions of a semi-simplicial complex and a complete semi-simplicial complex in 1950. The Eilenberg-Zilber notion of a complete
semi-simplicial complex is identical to our modern notion of a simplicial set K: it is
a sequence K0 , K1 , . . . of sets together with face maps ∂i : Kq → Kq−1 and degeneracy maps si : Kq → Kq+1 (0 ≤ i ≤ q) satisfying certain axioms; a semi-simplicial
complex is just a simplicial set without the degeneracy maps.
A word about changing terminology is in order. The term “complete semisimplicial complex” was awkward and was quickly abbreviated to “c.s.s. complex.”
During the 1950’s the term c.s.s. complex prevailed, although the short-lived term
“F D-complex” was also used in [EM54] and [D58]. Largely due to the influence
of John Moore, the adjective “complete” began to be omitted, starting with 1954,
while the notion of “semi-simplicial complex” languished in obscurity. By the early
1960’s the term “semi-simplicial set” had replaced “c.s.s. complex.” By the late
1960’s, even the prefix “semi” was dropped, influenced by the book [May]; since
then “simplicial set” has been the universally used term.
Returning to the early 1950’s, we mention two results which showed the power
of the new simplicial methods. The “Eilenberg-Zilber Theorem” was proven in
1953 [EZ53] as an application of c.s.s. complexes to products: the (simplicial)
map S(X × Y ) ' S(X) ⊗ S(Y ), implicitly defined by Alexander and Whitney in
1935, is a homotopy equivalence. In 1955, the homotopy theory of c.s.s. complexes
satisfying an extension condition was developed by Daniel Kan [Kan56]; a simplicial
set satisfying Kan’s extension condition is now called a Kan complex.
Homotopical algebra.
The homological study of simplicial abelian groups was launched by Eilenberg
and Mac Lane in [EM54], as part of their algebraic program to find the cohomology
of Eilenberg-Mac Lane spaces K(π, n). This program was analyzed with typical
thoroughness in the 1954/55 Seminaire Cartan [C55]. In exposés 18 and 19 of that
seminar, John Moore showed that every simplicial group K is a Kan complex, and
that one could compute its homotopy groups as the homology of a chain complex
N∗ of groups, where Nq ⊂ Kq is the intersection of kernels of all the face maps
except ∂q . The complex N∗ quickly became known as the Moore complex of K.
In 1956–57, A. Dold [D58] and D. Kan [Kan58] independently discovered that the
Moore complex provided an equivalence between the category of simplicial abelian
groups and the category of non-negative chain complexes of abelian groups. This
Dold-Kan correspondence was later codified in [DP]. Under the correspondence,
Moore’s result states that simplicial homotopy corresponds to homology. With this
correspondence at hand, simplicial techniques could be brought to bear on any
homological problem.
Dold and Puppe [DP] announced in 1958 that with simplicial methods one could
HISTORY OF HOMOLOGICAL ALGEBRA
29
define the derived functors of a non-additive functor T (say of modules); their detailed paper appeared in 1961. The key idea was that one could consider a projective
resolution P∗ of a module M as a simplicial module via the Dold-Kan correspondence. Since the notion of simplicial homotopy doesn’t involve addition, we may
take the homotopy groups of T (P∗ ) as the derived functors Li T (M ) of T . A variant
is obtained by placing M in degree n > 0; the derived functors Li T (M, n) are the
homotopy groups of T (P [n]), where the simplicial module P [n] corresponds to the
chain complex P∗ shifted n places. For example, the ith homology Hi (K(π, n); Z)
of an Eilenberg-Mac Lane space K(π, n) is just Li T (π, n) for the group ring functor
T (π) = Z[π].
It is possible to generalize the Dold-Puppe construction and define the left derived functors of any functor T from any category C to an abelian category, as
long as C is closed under finite limits and has enough projective objects. This observation evolved during the late 1960’s, finding voice in M. André’s book [A67],
Quillen’s book [Q67] on homotopical algebra, and in the later papers [A70, Q70].
In fact there are three standard constructions, which agree in reasonable situations.
André’s construction [A67] uses a subcategory of “acyclic models” in C. In the
category of functors on C, one finds a resolution T∗ → T which is aspherical on
the “model” objects. Then one defines Li T (A) to be πi T∗ (A), or Hi of the chain
complex associated to the simplicial module T∗ (A).
Quillen’s construction is simpler: one finds a simplicial “resolution” P∗ → A of
each A in C, and defines Li T (A) to be Hi T (P∗ ). The work comes in deciding what
a “resolution” is: P∗ should be cofibrant and P∗ → A should be an acyclic fibration
in the terminology of [Q67]. In many algebriac applications, fibrations are defined
by a relative lifting property, so all “relatively projective” objects are cofibrant.
During 1965–69, Barr and Beck [BB] developed the idea of cotriple resolutions as
a functorial way to obtain resolutions for computing nonabelian derived functors.
Suppose that there is a forgetful functor U : C → S with a left adjoint F . Then
the functor F U is called a cotriple, and the iterates Pi = (F U )i+1 (A) often form a
simplicial “resolution” P∗ → A. Again, one takes Li T (A) = Hi T (P∗ ).
Cohomology of commutative rings
In analogy with Hochschild’s (co)homology theory for associative algebras, it is
reasonable to ask for a (co)homology theory for commutative rings. Let k → A
be a map of commutative rings, and M an A-module. Then Hochschild’s group
H 1 (A; M ) is the A-module Derk (A, M ) of all derivations A → M which vanish on k
(as there are no inner derivations), H1 (A; M ) is M ⊗ ΩA/k and H 2 (A; M ) classifies
all associative k-algebra extensions B of A by M which are k-split, meaning that
B∼
= A ⊕ M as a k-module (this condition is obvious when k is a field). What was
wanted was a theory with the same H 1 and H1 , but such that H 2 was the group
Exalcommk (A, M ) classifying all commutative k-algebra extensions of A by M .
The functors H 1 and H 2 were first studied by P. Cartier [C56] in the case that
A = K is a field extension of k, and partially extended to commutative rings by
Nakai [N61]. In a 1961 course at Harvard, Grothendieck defined Exalcommk (A, M )
and constructed a 6-term cohomology sequence for k → A → B [EGA 0IV (18.4.2)].
When k is a field, Harrison [H62] used a subcomplex of the Hochschild complex
2
1
∗
= Exalcommk ,
= H 1 and Hharr
(A, M ) with Hharr
to define k-modules Hharr
30
CHARLES A. WEIBEL
equipped with a 9-term cohomology sequence. When k is perfect, and A is the
local ring (at some point) of a variety over k, Harrison proved the following two
2
results: (1) A is regular if and only if Hharr
(A, −) = 0, and (2) A is a complete
1
2
intersection if and only if dim Hharr (A, A/m) − dim Hharr
(A, A/m) = dim A.
The next step was taken in the 1964 paper [LS] by two Ph. D. students of
Tate, Lichtenbaum and Schlessinger. Let k be any commutative ring. For each
commutative ring map f : k → A, they defined a 3-term chain complex L· , called
the cotangent complex of f , and — for i = 0, 1, 2 — set Ti (A/k, M ) = Hi (L· ⊗ M ),
T i (A/k, M ) = H i Hom(L· , M ). When k is a field the T i (A/k, M ) agreed with
i+1
Harrison’s Hharr
(A, M ), and in general T 1 (A/k, M ) = Exalcommk (A, M ). The
vanishing of T 1 gave an Their infinitesimal criterion for A/k to be smooth, in terms
of the vanishing of T 1 (A/k), was later used by Grothendieck to great advantage in
[EGA IV.17]. If k is noetherian, R is a localization of k[x, . . . , y] and A = R/I,
they showed that T 2 (A/k, −) = 0 if and only if A is a complete intersection, i.e.,
I is defined by a regular sequence in R. Schlessinger’s thesis applied the T i to
deformation theory, while Lichtenbaum’s thesis was concerned with applications to
relative intersection theory.
In 1967, M. André [A67, A70, A74] and Quillen [Q70] discovered what we now call
André-Quillen cohomology. If k → A and M are as above, their groups D i (A/k, M )
agree with the Lichtenbaum-Schlessinger groups T i (A/k, M ) for i = 0, 1, 2. It
comes with a long exact sequence for k → A → B (generalizing Harrison’s) and
generalizations of the Lichtenbaum-Schlessinger results for smoothness and local
complete intersections. In this theory, the central role is played by a simplicial
A-module LA/k , called the cotangent complex of A relative to k, because of the
similarity (using the Dold-Kan correspondence) to the Lichtenbaum-Schlessinger
complex L· . This complex is well-defined in the derived category of chain complexes
of A-modules, and one has D i (A/k, M ) = H i HomA (LA/k , M ) and Di (A/k, M ) =
Hi (LA/k ⊗A M ).
Formally, the D i (A/k, M ) are the nonabelian derived functors of the functor
T (B) = Derk (B, M ) ∼
= HomA (A ⊗B ΩB/k ) on the category C of commutative kalgebras over A. According to the above prescription, the definition starts with an
acyclic simplicial resolution P∗ → A in C, and has D i (A/k, M ) = H i Derk (P∗ , M ).
Defining the simplicial A-module LA/k = A ⊗P ΩP/k , a little algebra yields the
above formulas.
Higher algebraic K-theory
In order to find a possible definition of the higher K-groups Kn (R) of a ring R,
Swan was led in 1968 to consider the nonabelian derived functors of the general
linear group GL on the category of rings [Sw70]. This required a slight generalization of derived functor, since the category of groups is not an abelian category. In
this context we have a functor G from a category C, such as the category of rings,
to the category of groups or sets.
Swan’s original construction followed André’s method, finding an acyclic resolution G∗ → GL in the functor category and setting Kn (R) = πn−2 G∗ (R) for n ≥ 2.
In 1969 Gersten gave a cotriple construction [G71], using the cotriple associated to
the forgetful functor from rings to sets, while both Keune [K71] and Swan [Sw72]
HISTORY OF HOMOLOGICAL ALGEBRA
31
gave constructions using free resolutions P∗ → R to define Kn (R) = πn−2 GL(P∗ )
for n ≥ 2. By 1970, Swan had proven [Sw72] that all three constructions yielded
the same functors Kn (R).
Historically, however, the important construction was given by Quillen in 1969
[Q71]. He showed how to modify the classifying space BGL(R) of GL(R) to obtain
a topological space BGL(R)+ with the same homology as BGL(R), and defined
Kn (R) = πn BGL(R)+ for n ≥ 1. The equivalence of Quillen’s topological definition
with the homological Swan-Gersten definition was established in 1972 by combining
partial results obtained by several authors [A73]. Since then the field of higher
algebraic K-theory has taken on a life of its own, but that is another story.
Hochschild and cyclic homology.
We have already described the 1945 development [Hh45] of Hochschild homology
of an algebra A over a field k. The next step was to let A be an algebra over an
arbitrary commutative base ring k. In his 1956 paper [Hh56], Hochschild began
a systematic study of exact sequences of R-modules which are k-split (split as
sequences of k-modules). This became part of a “relative” homological algebra
movement.
Hochschild, Kostant and Rosenberg showed in 1962 [HKR] that if A is smooth
of finite type over a field k, then there is a natural isomorphism Ω∗A/k ∼
= H∗ (A, A).
n+1
n
It follows that for such A there is an analogue d : ΩA → ΩA of de Rham’s operator for manifolds. Rinehart [R63] mimicked this construction for all algebras,
constructing a chain map B inducing an operator HHn (A, A) → Hn+1 (A, A). This
attempt to define an analogue of de Rham cohomology was before its time: twenty
years later Alain Connes [C85] as well as Feigin and Tsygan [T83, FT] would both
seize upon B and make it the foundation of cyclic homology, unaware of Rinehart’s
earlier work.
We end our quick tour by mentioning an important application, discovered by
Gerstenhaber in the 1964 paper [G64]. A deformation of an associative algebra A
is a k[[t]]-algebra structure on the k[[t]]-module A[[t]] whose product agrees modulo t with the given product on A. Reducing a deformation modulo t2 yields a
k-split algebra extension of A by A, so giving the “infinitesimal” part of the deformation is equivalent to giving an element of H 2 (A, A). Gerstenhaber showed
that there is a whole sequence of obstructions to deformations of A, lying in the
higher Hochschild cohomology groups. If A is smooth of finite type, the HochschildKostant-Rosenberg theorem implies that the obstructions belong to Ω∗A/k .
Cotor for coalgebras.
Hochschild homology was also involved in the early development of (differential
graded) coalgebras over a field. This field was heavily influenced by its applications
to topology, in part because the homology of a topological space X is a graded coalgebra, via the diagonal map H∗ (X) → H∗ (X × X) ∼
= H∗ (X) ⊗ H∗ (X). Moreover,
the normalized chain complex C∗ (X) is a differential graded coalgebra.
In 1956, J. F. Adams [A56] discovered a recipe for the homology of the loop
space ΩX when X is simply connected. To describe it, he considered C∗ (X) as a
differential graded coalgebra. Mimicking the Eilenberg-Mac Lane bar construction,
Adams defined a differential graded algebra F∗ , called the cobar construction, and
32
CHARLES A. WEIBEL
showed that H∗ (ΩX) ∼
= H∗ (F∗ ). This purely algebraic construction attracted the
attention of topologists to the algebraic structure of coalgebras and their comodules.
Now if C is a coalgebra one can define the cotensor product M C N of comodules M and N . Its right derived functors are called the cotorsion products
CotorC (M, N ) of M and N . In [EM66], Eilenberg and Moore defined and studied
the cotensor product over a DG coalgebra C = C∗ . Under mild flatness hypotheses, they constructed what we now call the “Eilenberg-Moore spectral sequence,”
C
which has E 2 equal to CotorHC
pq (H(M ), H(N )) and converges to Cotor (M, N ).
The importance of this is illustrated by the case when C is the normalized chain
complex of a simply connected topological space X, and M and N are the chain
complexes of spaces E and X 0 over X. If E → X is a Serre fibration, they prove
that CotorC (M, N ) is the homology of the fiber space E 0 = E ×X X 0 , so this provides a powerful method to calculate homology. Of course when X 0 and E are
contractible then E 0 ' ΩX, and they recover Adams’ cobar construction.
Eilenberg and Moore also studied the dual construction for tensor products of
differential graded modules M , N over a differential graded algebra R. In this case
the spectral sequence is E2pq = TorH(R) (H(N ), H(M )) ⇒ TorR (N, M ). Using the
cochain algebras in the above topological situation, Eilenberg and Moore proved
that H ∗ (E 0 ) ∼
= TorC ∗ (X) (C ∗ (E), C ∗ (X 0 )), so the spectral sequence converges to
H ∗ (E 0 ). This spectral sequence was described and studied in [S67] by Larry Smith,
who showed that this spectral sequence often collapsed.
Here is one application. Suppose that Y is simply connected and we take X =
Y × Y , with X 0 the diagonal copy of Y , and E the path space of Y . Then E 0 = ΩY
and if C ∗ (Y ) takes coefficients in a field k the Künneth formula yields C ∗ (X) '
C ∗ (Y ) ⊗ C ∗ (Y ). Since the Eilenberg-Moore spectral sequence collapses in this
case it yields an isomorphism between H ∗ (ΩY ), and the Hochschild cohomology
HH ∗ (C ∗ (Y ), k) of the differential graded algebra C ∗ (Y ).
Mac Lane Cohomology and Topological Hochschild Homology.
Let A be an associative ring and M an A-bimodule. As we have mentioned
above, the Hochschild cohomology group H 2 (A, A) only measures ring extensions
of A by M whose underlying abelian group is A ⊕ M . (One takes k to be Z.) In
order to measure all ring extensions of A by M , Mac Lane introduced what we now
call Mac Lane cohomology in the 1956 paper [M56]. One may naturally define a
differential graded ring Q = Q∗ (A) and an augmentation Q → A. By definition,
HM L∗ (A, M ) and HM L∗ (A, M ) are the Hochschild homology H∗ (Q, M ) and cohomology H ∗ (Q, M ). As required, ring extensions correspond to elements of the
group HM L2 (A, M ).
A variant for k-algebras and their extensions was invented in 1961 by U. Shukla
[Shuk], and is called Shukla homology. Shukla proved two comparison results: when
k is a field, Shukla homology recovers Hochschild homology; when k = Z, Shukla
homology recovers Mac Lane homology.
Both Mac Lane cohomology and Shukla homology were almost completely forgotten for thirty years, except for some calculations by Breen in [B78]. In 1991, an
innocuous paper by Jibladze and Pirashvili [JP91] proved that the Mac Lane homology of a ring A (and a module M ) is TorF
∗ (A⊗, M ⊗) in the functor category
F = F (A) of functors from the category of fin. gen. free A-modules to the category
HISTORY OF HOMOLOGICAL ALGEBRA
33
of A-modules. Similarly, the Mac Lane cohomology of A is ExtF (A⊗, M ⊗). This
was to lead to an unexpected connection to algebraic K-theory and manifolds.
In the late 1970’s, F. Waldhausen introduced a variant of algebraic K-theory,
which he called stable K-theory [W78]. His construction was designed to study the
homotopy theory of the diffeomorphism group of a manifold, and could be applied to
a ring spectrum A as well as ordinary rings. Following this lead in the early 1980’s,
M. Bökstedt [Bö] introduced a variant T HH∗ (A, A) of Hochschild homology for ring
spectra, called Topological Hochschild Homology. It is roughly obtained by replacing
rings by ring spectra and tensor products over k by smash products. In 1987,
Waldhausen announced that stable K-theory of A was isomorphic to T HH(A),
but the proof [DM94] took several years to appear.
Then in 1992, Pirashvili and Waldhausen [PW92] used the functor category
interpretation to prove that the Mac Lane homology group HM L(A, A) was the
same as T HH(A). This showed that homological algebra could be applied to
calculate the topological invariants of Waldhausen and Bökstedt. A new and active
field of research has been born out of this discovery.
Cyclic homology.
Cyclic homology arose simultaneously in several applications in the early 1980’s.
While studying applications of C ∗ -algebras to differential geometry in 1981,
Alain Connes was led to study Hochschild cochains which were invariant under
cyclic permutations of its arguments [C83, C85]. Realizing that such “cyclic”
cochains were preserved by the Hochschild coboundary gave him a new cohomology
theory, rapidly christened HC ∗ (A) and called the cyclic homology of A. Meanwhile,
Boris Tsygan [T83] was studying the homology of the Lie algebra gl(A) over a field
k of characteristic zero, and discovered that the Hopf algebra H∗ (gl(A); k) was the
tensor algebra on the homology groups Ki+ (A) of the complex of all Hochschild
chains invariant under cyclic permutation; the proof, and the cohomology version,
appeared in the paper [FT] by Feigin and Tsygan. This description of H∗ (gl(A); k)
was discovered independently by Loday and Quillen [LQ], and their paper made
the new subject of cyclic homology accessible to a large audience.
Both Connes and Tsygan discovered the following key structural sequence relating it to Hochschild homology; Rinehart’s operator [R63] is the composition BI.
S
B
I
· · · HCn+1 (A) −
→ HCn−1 (A) −→ Hn (A, A) −
→ HCn (A) · · ·
Using this sequence, Connes and others rediscovered and clarified the connection
with de Rham cohomology; for smooth algebras HCn (A) is a product of de Rham
n−1
cohomology groups, together with ΩnA/k /dΩA/k
.
In retrospect, cyclic homology had been hinted at in several places: pseudoisotopy theory [DHS], the homology of S 1 -spaces and in algebraic K-theory [L81].
Other applications soon arose. For example, Goodwillie showed in [G86] that the
cyclic homology (over Q) of a nilpotent ideal I is isomorphic to the algebraic Ktheory of I. Because of its diverse applications to other areas of mathematics, cyclic
homology became quickly established as a flourishing field in its own right.
It is impossible to give an accurate historical perspective on current developments. As tempting as it is, I shall refrain from doing so. Perhaps in fifty years the
history of homological algebra will be unrecognizable to us today. Let us hope so!
34
CHARLES A. WEIBEL
References
[A15]
[A26]
[A55]
[A56]
[A62]
[A67]
[A70]
[A73]
[A74]
[A82]
[AB56]
[AB58]
[AB59]
[ANT]
[B28]
[B35]
[B34]
[B40]
[B42]
[B55]
[B61]
[B63]
[B78]
[B81]
[BB]
[BBD]
[BBG]
[Betti]
[BGS]
[BN]
[Boll]
[Bö]
J. W. Alexander, A proof of the invariance of certain constants in Analysis Situs, Trans.
AMS 16 (1915), 148–154.
J. W. Alexander, Combinatorial Analysis Situs, Trans. AMS 28 (1926), 301–329.
M. Auslander, On the dimension of modules and algebras. III. Global dimension, Nagoya
Math. J. 9 (1955), 67–77.
J. F. Adams, On the cobar construction for coalgebras, Proc. Nat. Acad. Sci. 42 (1956),
409–412.
M. Artin, Grothendieck Topologies, Mimeographed seminar notes, Harvard Univ., 1962.
M. André, Méthode simplicial en algèbre homologique et algèbre commutative, Lecture
Notes in Math. vol. 32, Springer Verlag, 1967.
M. André, Homology of simplicial objects, Proc. Sympos. Pure Math., vol. XVII, 1970,
pp. 15–36.
D. Anderson, Relationship among K-theories, Lecture Notes in Math. vol. 341, Springer
Verlag, 1973, pp. 57–72.
M. André, Homologie des algèbres commutatives, Springer Verlag, 1974.
D. Anick, A counterexample to a conjecture of Serre, Annals of Math. 115 (1982), 1–33.
M. Auslander and D. Buchsbaum, Homological dimension in local rings, Trans. AMS
85 (1957), 390–405; Homological dimension in Noetherian rings, Proc. Nat. Acad. Sci.
42 (1956), 36–38.
M. Auslander and D. Buchsbaum, Codimension and Multiplicity, Annals of Math. 68
(1958), 625–657.
M. Auslander and D. Buchsbaum, Unique factorization in regular local rings, Trans.
AMS 45 (1959), 733–734.
E. Artin, C. Nesbitt and R. Thrall, Rings with Minimim Condition, Univ. Michigan
Press, 1944.
R. Brauer, Untersuchungen über die arithmetischen Eigenschaften von Gruppen linearer
Substitutionen. I., Math. Z. 28 (1928), 677–696.
R. Brauer, Sur les invariants intégraux des variétés répresentatives des groupes de Lie
simples clos, C. R. Acad. Sci. (Paris) 201 (1935), 419–421.
R. Baer, Erwiterung von Gruppen und ihren Isomorphismen, Math. Zeit. 38 (1934),
375–416.
R. Baer, Abelian groups that are direct summands of every containing abelian group,
Bull AMS (1940), 800–806.
N. Bourbaki, Éléments de Mathématique. Part I., Livre II, Algèbre ch. 2, Hermann,
1942.
D. Buchsbaum, Exact categories and Duality, Trans. AMS 80 (1955), 1–34.
H. Bass, Projective modules over algebras, Annals Math. 73 (1961), 532–542.
H. Bass, On the ubiquity of Gorenstein rings, Math. Zeit. 82 (1963), 8–28.
L. Breen, Extensions du groupe additif, Publ. I.H.E.S. 48 (1978), 39–125.
J. M. Boardman, Conditionally convergent spectral sequences, preprint (1981).
M. Barr and J. Beck, Homology and standard constructions, Lecture Notes in Math. vol.
80, Springer Verlag, 1969.
A. Beilinson, J. Bernstein and P. Deligne, Faisceaux pervers, Astérisque 100 (1982).
I. Bernšteı̆n, I. Gelfand and S. Gelfand, Algebraic vector bundles on ¶n and problems of
linear algebra, Funkt. Anal. i Prilozhen. 12 (1978), 66–67.
E. Betti, Sopra gli spazi di un numero qualunque di dimensioni, Ann. Mat. pura appl.
2/4 (1871), 140–158.
A. Beı̆linson, V. Ginsburg and V. Schechtman, Koszul duality, J. Geom. Phys. 5 (1988),
317–350.
R. Brauer and E. Noether, Über minimale Zerfällungskörper irreduzibler Darstellungen,
Sitzungsberichte Preußischen Akad. Wissen. (1927), 221–228.
M. Bollinger, Geschichtliche Entwicklung des Hologiebegriffs, Archive for History of Exact Sciences 9 (1972), 94–170.
M. Bökstedt, Topological Hochschild Homology, preprint 1985.
HISTORY OF HOMOLOGICAL ALGEBRA
[BS]
[C28]
[C29]
[C32]
[C35]
[C48]
[C50]
[C55]
[C56]
[C83]
[C85]
[CE]
[ChE]
[CPS]
[D58]
[D74]
[DHS]
[Dieu]
[DM94]
[DP]
[dR]
[E44]
[E49]
[Eck]
[EGA]
[EM42]
[EM43]
[EM45]
[EM47]
35
A. Borel and J.-P. Serre, Le théorème de Riemann-Roch, Bull. Soc. Math. France 86
(1958), 97–136.
E. Cartan, Sur les nombres de Betti des espaces de groupes clos, C. R. Acad. Sci. (Paris)
187 (1928), 196–198.
E. Cartan, Sur les invariants intégraux de certains espaces homogènes clos et les propriétés topologique de ces espaces, Ann. Soc. Polon. Math. 8 (1929), 181–225.
E. Čech, Théorie générale de l’homologie dans un espace quelconque, Fundamenta Math.
19 (1932), 149–183.
, Les groupes de Betti d’un complex infini, Fundamenta Math. 25 (1935), 33–44.
H. Cartan, Séminaire Henri Cartan de topologie algébrique, 1948/1949., Secrétariat
mathématique, Paris, 1955.
H. Cartan, Séminaire Henri Cartan de topologie algébrique, 1950/1951., Cohomologie
des groupes, suite spectrale, faisceaux, Secrétariat mathématique, Paris, 1955.
H. Cartan, Séminaire Henri Cartan, 1954/1955., Algèbre de Eilenberg-Mac Lane et
homotopie, Secrétariat mathématique, Paris, 1955.
P. Cartier, Dérivations dans les corps, Exposé13, Seminaire Cartan et Chevalley 1955/56,
Secrétariat mathématique, Paris, 1956.
A. Connes, Cohomologie cyclique et foncteurs Extn , C. R. Acad. Sci. Paris 296 (1983),
953–958.
A. Connes, Non-commutative differential geometry, Publ. I.H.E.S. 62 (1985), 257–360.
H. Cartan and S. Eilenberg, Homological Algebra, Princeton U. Press, 1956.
C. Chevalley and S. Eilenberg, Cohomology Theory of Lie Groups and Lie Algebras,
Trans. AMS 63 (1948), 85–124.
E. Cline, B. Parshall, and L. Scott, Derived categories and Morita theory, J. Algebra
104 (1986), 397–409.
A. Dold, Homology of symmetric products and other functors of complexes, Annals of
Math. 68 (1958), 54–80.
P. Deligne, La conjecture de Weil. I, Publ. Math. I.H.E.S. 43 (1974), 273–307.
W. Dwyer, W.-C. Hsiang and R. Staffeldt, Pseudo-isotopy and invariant theory, Topology 19 (1980), 367–385.
J. Diedonné, A History of Algebraic and Differential Topology 1900–1960, Birkhäuser,
1989.
B. Dundas and R. McCarthy, Stable K-theory and topological Hochschild homology,
Annals Math. 140 (1994), 685–701.
A. Dold and D. Puppe, Homologie nicht-additiver Funktoren; Anwendungen, Ann. Inst.
Fourier (Grenoble) 11 (1961), 201–312; Non-additive functors, their derived functors,
and the suspension homomorphism, Proc. Nat. Acad. Sci. 44 (1958), 1065–1068.
G. de Rham, Sur l’analysis situs des variétés à n dimensions, J. Math. Pures Appl. 10
(1931), 115–200; C. R. Acad. Sci. (Paris) 188 (1929), 1651–1652.
S. Eilenberg, Singular homology theory, Annals of Math. 45 (1944), 407–447.
S. Eilenberg, Topological methods in abstract algebra, Bull. AMS 55 (1949).
B. Eckmann, Der Cohomologie-Ring einer beliebigen Gruppe, Comment. Math. Helv.
18 (1945–46), 232–282.
A. Grothendieck and J. Dieudonné, Éléments de Géométrie Algèbrique, Part I: 4 (1960);
Part II: 8(1961); Part III: 11 (1961), 17 (1963); Part IV: 20 (1964), 24 (1965), 28 (1966),
32 (1967), Publ. I. H. E. S..
S. Eilenberg and S. Mac Lane, Group extensions and homology, Annals of Math. 43
(1942), 757–831.
S. Eilenberg and S. Mac Lane, Relations between homology and homotopy groups, Proc.
Nat. Acad. Sci. 29 (1943), 155–158; Relations between homology and homotopy groups
of spaces, Annals of Math. 46 (1945), 480–509.
S. Eilenberg and S. Mac Lane, Natural isomorphisms in group theory, Proc. Nat. Acad.
Sci. 28 (1942); General theory of natural equivalences, Trans. AMS 58 (1945), 231–294.
S. Eilenberg and S. Mac Lane, Cohomology theory in abstract groups. I, Annals of Math.
48 (1947), 51–78.
36
CHARLES A. WEIBEL
[EM54] S. Eilenberg and S. Mac Lane, On the groups H(Π, n). II, Annals of Math. 60 (1954),
49–139.
[EM62] S. Eilenberg and J. Moore, Limits and spectral sequences, Topology 1 (1962), 1–23.
[EM66] S. Eilenberg and J. Moore, Homology and Fibrations. I. Coalgebras, cotensor product
and its derived functors, Comment. Math. Helv. 40 (1966), 199–236.
[ES45] S. Eilenberg and N. Steenrod, Axiomatic approach to homology theory, Proc. Nat. Acad.
Sci. USA 31 (1945), 117–120.
[ES]
S. Eilenberg and N. Steenrod, Foundations of algebraic topology, Princeton U. Press,
Princeton, 1952.
[EZ50] S. Eilenberg and J. Zilber, Semi-simplicial complexes and singular homology, Annals of
Math. 51 (1950), 499–513.
[EZ53] S. Eilenberg and J. Zilber, On products of complexes, Amer. J. Math. 75 (1953), 200–204.
[F36]
H. Fitting, Über den Zusammenhang zwischen dem Begriff der Gleichartigkeit zweier
Ideale und dem Äquivalenzbegriff der Elementarteilertheorie, Math. Annalen 112 (1936),
572–582.
[F46]
H. Freudenthal, Der Einfluss der Fundamental Gruppe auf die Bettischen Gruppen,
Annals of Math. 47 (46), 274–316.
[F64]
J. P. Freyd, Abelian Categories, Harper & Row, 1964.
[FT]
B. Feigin and B. Tsygan, Cohomology of Lie algebras of generalized Jacobi matrices,
Funkt. Anal. Appl. 17 (1983 86–87); Funct. Anal. Appl. 17 (1983), 153–155.
[G57]
A. Grothendieck, Sur quelques points d’algèbre homologique, Tohoku Math. J. 9 (1957),
119–221.
[G62]
P. Gabriel, Des catégories abéliennes, Bull. Soc. Math. France 90 (1962), 323–428.
[G64]
M. Gerstenhaber, On the deformation of rings and algebras, Annals of Math. 79 (1964),
59–103; II, Annals of Math. 84 (1966), 1–19.
[G65]
J. Giraud, Cohomologie non abélienne, C. R. Acad. Sci. (Paris) 260 (1965), 2666–2668.
[G67]
A. Grothendieck, Local Cohomology, Lecture Notes vol. 41, Springer Verlag, 1967.
[G71]
S. Gersten, On the functor K2 , J. Algebra 17 (1971), 212–237.
[G86]
T. Goodwillie, Relative algebraic K-theory and cyclic homology, Annals Math. 24 (1986),
347–402.
[GAGA] J.-P. Serre, Géométrie algébrique et géométrie analytique, Ann. Inst. Fourier (Grenoble)
6 (1955–56), 1–42.
[GFGA] A. Grothendieck, Fondements de la géométrie algébrique, Extraits du Séminaire Bourbaki, 1957–1962, Secr’etariat mathématique, Paris, 1962.
[Gir63] J. Giraud, Analysis situs, Séminaire Bourbaki 1962/63, No. 256, 1964.
[Gode] R. Godement, Topologie algébrique et théorie des faisceaux, Hermann, 1958.
[H38]
M. Hall, Group rings and extensions, I, Annals of Math. 39 (1938), 220–234.
[H56]
M. Harada, Note on the dimension of modules and algebras, J. Inst. Poly. Osaka Univ.
Ser. A 7 (1956), 17–27.
[H58]
A. Heller, Homological Algebra in Abelian Categories, Annals of Math. 68 (1958), 484–
525.
[H62]
D. K. Harrison, Commutative algebras and cohomology, Trans AMS 104 (1962), 191–
204.
[H66]
R. Hartshorne, Residues and Duality, Springer Lecture Notes vol. 20, Springer Verlag,
1966.
[Hf41]
H. Hopf, Über die Topologie der Gruppen-Mannigfaltigkeiten und ihre Verallgemeinerungen, Annals of Math. 42 (1941), 22-52.
[Hf42]
H. Hopf, Fundamentalgruppe und zweite Bettische Gruppe, Comment. Math. Helv. 14
(1941–42), 257–309; announced, Relations between the fundamental group and the second
Betti Group, Lectures in Topology, Ann Arbor, 1942, pp. 315–316.
[Hf44]
H. Hopf, Über die Bettischen Gruppen, die zu einer beliebigen Gruppe gehören, Comment. Math. Helv. 17 (1944–45), 39–79.
[Hh42] G. Hochschild, Semisimple algebras and generalized derivations, Amer. J. Math. 64
(1942), 677–694.
[Hh45] G. Hochschild, On the cohomology groups of an associative algebra, Annals of Math. 46
(1945), 58–67.
HISTORY OF HOMOLOGICAL ALGEBRA
[Hh50]
[Hilb]
37
G. Hochschild, Local class field theory, Annals of Math. 51 (1950), 331–347.
D. Hilbert, Über die Theorie der Algebraischen Formen, Math. Annalen 36 (1890),
473–534.
[HKR] G. Hochschild, B. Kostant and A. Rosenberg, Differential forms on regular affine algebras, Trans. AMS 102 (1962), 383–408.
[HK71] J. Herzog and E. Kunz, Der kanonische Modul eines Cohen-Macauley-Rings, Lecture
Notes vol. 238, Springer Verlag, 1971.
[HR76] M. Hochster and J. Roberts, The purity of the Frobenius and local cohomology, Advances
in Math. 21]yr1976, 117–172..
[HS53] G. Hochschild and J.-P. Serre, Cohomology of group extensions, Trans AMS 74 (1953),
110–134; Cohomology of Lie algebras, Annals of Math. 57 (1953), 591—603.
[HN52] G. Hochschild and T. Nakayama, Cohomology in Class Field Theory, Annals of Math.
55 (1952), 348–366.
[Hö]
O. Hölder, Die Gruppen der Ordnungen p3 , pq 2 , pqr, p4 , Math. Ann. 43 (1893), 301–
412.
[HStm] P. Hilton and U. Stammbach, A Course in Homological Algebra, Springer Verlag, 1971.
[Hu36] W. Hurewicz, Beitrage zur Topologie der Deformationen. IV. Asphärische Räume, Proc.
Akad. Amsterdam 39 (1936), 215–224.
[Hu41] W. Hurewicz, On Duality Theorems, abstract 47-7-329, Bull. AMS 47 (1941), 562–563.
[J37]
N. Jacobson, Abstract derivation and Lie algebras, Trans AMS 42 (1937), 206–224.
[JP91] M. Jibladze and T. Pirashvili, Cohomology of algebraic theories, J. Algebra 137 (1991),
253–296.
[K23]
H. Künneth, Über die Bettischen Zahlen einer Produktmannigfaltigkeit, Math. Ann. 90
(1923), 65–85; Über die Torsionszahlen von Produktmannigfaltigkeiten, Math. Ann. 91
(1924), 125–134.
[K47]
J.-L. Koszul, Sur les opérateurs de dérivation dans un anneau, C. R. Acad. Sci. (Paris)
224 (1947), 217–219; Sur l’homogie des espaces homogènes, 477-479.
[K50]
J.-L. Koszul, Homologie et cohomologie des algèbres de Lie, Bull. Soc. Math. France 78
(1950), 65–127.
[K52]
I. Kaplansky, Modules over Dedekind rings and Valuation rings, Trans. AMS 72 (1952),
327–340.
[KT55] Kawada and J. Tate, On the Galois cohomology of unramified extensions of function
fields in one variable., Amer. J. Math. 77 (1955), 197–217.
[Kan56] D. Kan, Abstract Homotopy. III, IV, Proc. Nat. Acad. Sci. 42 (1956), 419–421, 542–544.
[Kan58] D. Kan, On functors involving c.s.s. complexes, Trans. AMS 87 (1958), 330–346.
[K58]
I. Kaplansky, Projective Modules, Annals of Math. 68 (1958), 372–377.
[K59]
I. Kaplansky, Homological dimension of rings and modules, Univ. of Chicago Dept. of
Math., 1959, reprinted as part III of “Fields and Rings,” Univ. of Chicago Press, 1969.
[K71]
F. Keune, Algèbre homotopique et K-théorie algébrique, C.R. Acad. Sci. Paris 273
(1971), A592–A595; Derived functors and algebraic K-theory, Lecture Notes in Math.
vol. 341, Springer Verlag, 1973, pp. 166–178.
[KP]
J. Kelley and E. Pritcher, Exact homomorphism sequences in homology theory, Annals
of Math. 48 (1947), 682–709.
[L33]
S. Lefschetz, On singular chains and cycles, Bull. AMS 39 (1933), 124–129.
[L46a]
J. Leray, L’anneau d’homologie d’une représentation, C. R. Acad. Sci. (Paris) 222
(1946), 1366–1368.
[L46b]
, Stucture de l’anneau d’homologie d’une représentation, C.R. Acad. Sci. (Paris)
222 (1946), 1419–1422.
[L50]
, L’anneau spectral et l’anneau filtré d’homologie d’un espace localement compact
et d’une application continue, J. Math. Pures Appl. 29 (1950), 1–139.
[L60]
S. Lubkin, Imbedding of abelian categories, Trans. AMS 97 (1960), 410–417.
[L81]
J.-L. Loday, Symboles en K-théorie algébrique supérieure, C. R. Acad. Sci. Paris 292
(1981), 863–866.
[LQ]
J.-L. Loday and D. Quillen, Homologie cyclique et homologie de l’algèbre de Lie des
matrices, C. R. Acad. Sci. Paris 296 (1983), 295–297; Cyclic homology and the Lie
algebra homology of matrices, Comment. Math. Helv. 59 (1984), 569–591.
38
[LS]
CHARLES A. WEIBEL
S. Lichtenbaum and M. Schlessinger, The cotangent complex of a morphism, Trans. AMS
128 (1967), 41–70.
[LT]
S. Lang and J. Tate, Principal Homogeneous Spaces over Abelian Varieties, Amer. J.
Math 80 (1958), 659–684.
[Lyn]
R. Lyndon, The cohomology theory of group extensions, Duke Math. J. 15 (1948), 271–
292.
[M29]
L. Mayer, Über Abstakte Topologie. I und II, Monatsh. Math u. Physik 36 (1929), 1–42,
219–258.
[M48]
S. Mac Lane, Groups, categories and duality, Proc. Nat. Acad. Sci. 34 (1948); Duality
for groups, Bull. AMS 56 (1950), 485–516.
[M52]
W. Massey, Exact couples in algebraic topology, Annals of Math. 56 (1952), 363–396.
[M56]
S. Mac Lane, Homologie des anneau et des modules, Colloque de topologie algèbrique,
Georges Thon, Liège, 1956, pp. 55–80.
[M58]
E. Matlis, Injective modules over noetherian rings, Pac. J. Math. 8 (1958), 511–528.
[M62]
J. Milnor, On axiomatic homology theory, Pacific J. Math. 12 (1962), 337–341.
[M88]
S. Mac Lane, Group extensions for 45 years, Math. Intelligencer 10 (1988), 29–35..
[M89]
S. Mac Lane, The Applied Mathematics Group at Columbia in World War II, A Century
of Mathematics in America (P. Duren, ed.), vol. III, AMS, 1989, pp. 495–515.
[May]
J. P. May, Stable algebraic topology, The History of Topology (I. M. James, ed.), 1998.
[MHC] W. Massey, A history of cohomology theory, The History of Topology (I. M. James, ed.),
1998.
[ML]
S. Mac Lane, Homology, Springer-Verlag, 1963.
[Mit]
B. Mitchell, Amer. J. Math.
[N25]
E. Noether, Ableitung der Elementarteilertheorie aus der Gruppentheorie, Nachrichten
der 27 Januar 1925, Jahresbericht Deutschen Math. Verein. (2. Abteilung) 34 (1926),
104.
[N58]
M. Nagata, A general theory of algebraic geometry over Dedekind domains. II. Separably
generated extensions and regular local rings, Amer. J. Math. 80 (1958), 382–420.
[N61]
Y. Nakai, On the theory of differentials in commutative rings, J. Math. Soc. Japan 13
(1961), 63–84.
[O68]
B. Osofsky, Homological dimension and the continuum hypothesis, Trans. AMS 132
(1968), 217–230.
[Poin]
H. Poincaré, Œvres, edited by R. Garnier and J. Leray, vol. 6, Gauthiers-Villars, Paris,
1953.
[P35]
L. Pontrjagin, Sur les nombres de Betti des groupes de Lie, C. R. Acad. Sci. (Paris)
200 (1935), 1277–1280; Homologies in Compact Lie Groups, Mat. Sbornik 6 (1939),
389–422.
[PW92] T. Pirashvili, F. Waldhausen, Mac Lane homology and topological Hochschild homology,
J. Pure Appl. Algebra 82 (1992), 81-98.
[Q67]
D. Quillen, Homotopical Algebra, Lecture Notes vol. 43, Springer Verlag, 1967.
[Q70]
D. Quillen, On the (co)homology of commutative rings, Proc. Sympos. Pure Math.,
vol. XVII, 1970, pp. 65–87.
[Q71]
D. Quillen, Cohomology of groups, Proc. I. C. M. Nice 1970, 1971, pp. 47–51.
[Q76]
D. Quillen, Projective modules over polynomial rings, Invent. Math. 36 (1976).
[Riem] B.Riemann, Gesammelte Mathematische Werke und Wissenschaftlicher Nachlass, edited
by H. Weber, Teubner, Leipzig, 1876.
[R61]
J. E. Roos, Sur les foncteurs dérivés de lim, C. R. Acad. Sci. (Paris) 252 (1961), 3702–
3704.
[R63]
G. Rinehart, Differential forms on general commutative algebras, Trans.AMS 108 (1963),
195–222.
[R79]
J. E. Roos, Relations between the Poincaré-Betti series of loop spaces and of local rings,
Lecture Notes in Math. vol. 740, Springer Verlag, 1979, pp. 285–322.
[Rot]
J. Rotman, Notes on Homological Algebra, Van Nostrand, 1970; An Introduction to
Homological Algebra, Academic Press, 1979.
[S04]
I. Schur, Über die Darstellungen der endlichen Gruppen durch gegebene lineare Substitutionen, J. Reine Angew. Math. 127 (1904), 20–50.
HISTORY OF HOMOLOGICAL ALGEBRA
[S26]
[S36]
[S40]
[S67]
[S70]
[S71]
[S76]
[S88]
[Se50]
[Se51]
[Se55]
[Se56]
[Se58]
[SeCG]
[SeM]
[SGA4]
[SGA6]
[Shuk]
[Sw59]
[Sw62]
[Sw70]
[Sw72]
[T52]
[T54]
[T57]
[T83]
[V]
[VA]
[W36]
[W37]
[W38]
39
O. Schreier, Über die Erweiterungen von Gruppen. I, Monatsh. Math. u. Phys. 34 (1926),
165–180.
N. Steenrod, Universal Homology Groups, Amer. J. Math. 58 (1936), 661-701.
N. Steenrod, Regular cycles of compact metric spaces, Annals of Math. 41 (1940), 833–
851.
L. Smith, Homological algebra and the Eilenberg-Moore spectral sequence, Trans. AMS
129 (1967), 58–93.
R. Sharp, Local cohomology theory in commutative algebra, Quart. J. Math. 21 (1970),
425–434.
R. Sharp, On Gorenstein modules over a complete Cohen-Macaulay local ring, Quart.
J. Math. 22 (1971), 425–434.
A. Suslin, Projective modules over polynomial rings are free, Dokl. Akad. Nauk SSSR
229 (1976), 1063–1066; Soviet Math. Dokl. 17 (1976), 1160–1164.
N. Spaltenstein, Resolutions of unbounded complexes, Compositio Math. 65 (1988), 121–
154.
J.-P. Serre, Cohomologie des extensions de groupes, C. R. Acad. Sci. (Paris) 231 (1950),
643–646.
J.-P. Serre, Homologie singulière des espaces fibrés, Annals of Math. 54 (1951), 425–505.
J.-P. Serre, Faisceaux algébrique cohérents, Annals of Math. 61 (1955), 197–278.
J.-P. Serre, Sur la dimension homologique des anneaux et des modules noethériens,
Proc. Sympos. Algebraic Number Theory 1955, Science Council of Japan, Tokyo, 1956,
pp. 175–189.
J.-P. Serre, Modules projectifs et espaces fibrés à fibre vectorielle, Exposé 23, Seminaire
P. Dubreil 1957/58, Secrétariat mathématique, Paris, 1958.
, Cohomologie Galoisienne, Lecture Notes in Math. vol. 5, Springer Verlag, 1964.
, Algèbre Locale Multiplicités, Mimeographed Notes, I.H.E.S., 1957; Lecture
Notes in Math. vol. 11, Springer Verlag, 1965.
M. Artin, A. Grothendieck and J.-L. Verdier, Théorie des Topos et Cohomologie Etale
des Schémas, Lecture Notes in Math. vols. 269, 270, 305, Springer Verlag, 1972–73.
P. Berthelot, A. Grothendieck and L. Illusie, Théorie des Intersections et Théorème de
Riemann-Roch, Lecture Notes in Math. vol. 225, Springer Verlag, 1971.
U. Shukla, Cohomologie des algèbres associatives, Ann. Sci. Éc. Norm. Sup. 78 (1961),
163–209.
R. Swan, Projective modules over finite groups, Bull. AMS 65 (1959), 365–367.
R. Swan, Vector bundles and projective modules, Trans. AMS 105 (1962), 264–277.
R. Swan, Nonabelian homological algebra and K-theory, Proc. Sympos. Pure Math.,
vol. XVII, 1970, pp. 88–123.
R. Swan, Some relations between higher K-functors, J. Alg. 21 (1972), 113–136.
J. Tate, The higher dimensional cohomology groups of class field theory, Annals of Math.
56 (1952), 294–297.
J. Tate, The cohomology groups of algebraic number fields, Proc. I. C. M. Amsterdam
1954, North Holland, 1954, pp. 66–67.
J. Tate, Homology of noetherian rings and local rings, Illinois J. Math. 1 (1957), 14–27.
B. Tsygan, Homology of matrix Lie algebras over rings and the Hochschild homology.
(Russian), Uspekhi Mat. Nauk 38 (1983), 217–218; Russian Math. Surveys 38 (1983),
198–199.
J.-L. Verdier, Catégories dérivées, état 0, SGA 4 12 , Lecture Notes in Math. vol. 569,
Springer Verlag, 1977.
O. Veblen and J. Alexander, Manifolds of n dimensions, Annals of Math. 14 (1913),
163–178.
J. H. C. Whitehead, On the decomposition of an infinitesimal group, Proc. Camb. Phil.
Soc. 32 (1936), 229–237; Certain equations in the algebra of a semi-simple infinitesimal
group, Quart. J. Math. 8 (1937), 220–237.
H. Whitney, On products in a complex, Proc. Nat. Acad. Sci. 23 (1937), 285–291; Annals
of Math. 39 (1938), 397–432.
H. Whitney, Tensor products of Abelian Groups, Duke Math. J. 4 (1938), 495–528.
40
CHARLES A. WEIBEL
[W78]
[Y54]
F. Waldhausen, Algebraic K-theory of spaces I, Proc. Symp. Pure Math., vol. 32, 1978,
pp. 35–60.
N. Yoneda, On the homology theory of modules J. Fac. Sci. Tokyo 7 (1954), 193–227.
Department of Mathematics, Rutgers University, U.S.A.
Download